1965 11_#27_Part_2 11 #27 Part 2

1965-11_#27_Part_2 1965-11_%2327_Part_2

User Manual: 1965-11_#27_Part_2

Open the PDF directly: View PDF .
Page Count: 199

Download
Open PDF In Browser	View PDF

AFIPS
CONFERENCE
PROCEEDINGS

,,
THEIR IMPACT
ON SOCIETY

(Volume 27, Part 2
1965 Fall Joint
Computer Conf.rence)

1967
THOMPSON BOOK COMPANY
Washington, D.C.
ACADEMIC PRESS, INC.
London

The ideas and opinions expressed herein are solely "those of the authors and
are not necessarily representative of or endorsed by the 1965 Fall Joint Computer
Conference Committee or the American Federation of Information Processing
Societies.

Library of Congress Catalog Card Number 55-44701
Thompson Book Company
383 National Press Building
Washington, D. C.

1967 by the American Federation of Information Processing Societies, 345 E.
47th St., New York, New York 10017. All rights reserved. This book, or parts
thereof, may not be reproduced in any form without permission of the publishers.

Sole distributors throughout the world, with the exception of
North and South America:
Academic Press, Inc.
Berkeley Square
London, WI, England

Contents
Foreword .............................................. .
Preface ................................................ .
Conference Committees .................................. .
Harry Goode Memorial Award ............................ .
The mighty man-computer team ............................ .
The computer and our changing society .. . . . . . . . . . . . ......... .
Computers and education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Computers: The physical sciences and medicine . . . . . . . . . . . . . . . .
Impact of computers on retailing . . . . . . . . . . . . . . . . . . . . . . . . . . ...
The application of computers to domestic and international trade. . . .
The role of computers in space exploration . . . . . . . . . . . . . . . . . . . .
The impact of computers on the government .................. .
Communications, computers and people ...................... .
The impact of computers on urban transportation .. . . . . . . . . . . . .

I. Rhodes
S. Ramo
R. Gerard
I. Maloney, Ir.
C. McBrier
W. Merkin and R. Long
C. Gates and W. Pickering
I. Ward
P. Baran
K. Schlager

PANEL DISCUSSION RECORDS
The computer industry in the buyer's market

I. Ricca, I. Eckert,
W. Gallagher, R. Hubner
and R. Schmidt

The overseas computer market ............................. .

M. Mapes, Jr., D. Orr,

J. Miles, N. Ream and
T. Thau
The future of electromechanical mass storage ... . . . . . . . . . . . . . . .

W. Farrand, R. Franklin,
N. Hardy, R. Graham,
M. Eyster, W. Broderick,
F. Lohan, D. Sampson,
A. Sh~gart, I. Wieselman

Promising avenues for computer research . . . . . . . . . . . . . . . . . . . . . .

R. Rice, K. Uncapher, T. Steel
and L. Hobbs

LATE PAPERS NOT APPPEARING IN VOLUME 27, PART I
A high speed thin-film memory: Its design and development . . . . . .
Efficiency and management of a computing system.. ........... .
Use of digital computers in basic mathematic courses . . . . . . . . . . . . .

Q. Simkins
H. Huskey
W. Marsland, Ir.

ERRATA: Final Version of Papers Appearing in Preliminary Draft Form
in Part 1, Volume 27
MATHLAB: A' program for on-line machine assistance in symbolic
computations ....................... .................... .
A time-and memory-sharing executive program for quick response,
on-line applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .'. . .
Interactive, machine-language programming" . . . . . . . . . . . . . . . . . . . .
An integrated computer system for engineering problem solving. . . . .
ERRATA TO VOLUME 27, PART I

C. Engelman

I. Forgie
B. Lampson
D. Roos

Foreword
This second volume of the Proceedings of the 1965 Fall Joint Computer Conference is an attempt to capture the spirit of innovation that was the hallmark of
the Conference held in Las Vegas, Nevada, November 30 - December 1, 1965.
Part I of Volume 27, containing the formal papers presented in the Technical
Sessions, was distributed at the Conference. The first volume included the traditional papers covering recent advances in hardware and software as well as a
number of presentations focusing attention on new applications and problems of
management in information processing.
From its inception, the 1965 FJCC attempted to provide the opportunity for
professional communication at every leveI. To this end the conference included
a number of panel sessions organized to explore new areas of activities in the
field of information processing. In a more radical departure, the structure of the
traditional program of the conference was altered to focus the attention of the
entire conference on the profound influence that computing is exerting on all,
facets of American society. In place of the usual keynote speaker, the Conference
Committee invited nine outstanding speakers from disciplines outside of the traditional definition of data processing to explore the question of how computers are
affecting the world around us. The speakers from government, education, and
industry who honored our platform on the day-long session that closed the Conference, were not casual observers remote from the changing world of computer
technologies, but rather men of stature that knew and used computers in their
diverse fields.
This volume brings together the excitement of these phases of the conference,
the. edited results of significant panel sessions, and the thoughtful remarks of the
invited speakers. Also honored in the record are two pioneers in the field of computers and information processing, George Robert Stibitz and Konrad Zuse,
recipients of the second Harry Goode Memorial Award. It is fitting that AFIPS
should record in a permanent record the tributes and responses given in connection with this award. Although Miss Ida Rhodes was personally unable to attend
the conference due to illness in her immediate family, the complete text of her
Luncheon Speech is made a part of this volume.
A post-conference volume also is an opportunity to give public recognition
to a dedicated Conference Committee. The Committee was responsible for a
successful conference that set records in attendance and opened new vistas in
the information processing field. The names of the members of the several committees are here recorded in recognition of a job well done.
Special recognition goes to Samuel Nissim and Sei Shohara, who together
made this second volume possible.
ROBERT W. RECTOR
General Chairman
1965 Failloint Computer Conference
v

Preface
made available in advance of the conference. This permitted devoting the session to a discussion of these
papers rather than a formal presentation of them.
Despite the fact that the preprints were sent out late,
and that a few session chairmen and authors did not
really implement the Discuss-Only concept well, most
of the conference attendees polled were enthusiastic
about the results. It must be emphasized, though, that
only the concept of Discuss-Only sessions and its utility
have been proven. The many flaws that were discovered
in its implementation still require a significant amount
of "polishing." It should be noted that all papers in the
Preprints are included in Part 1 or 2 of the Conference
Proceedings.
Voice prints (on magnetic tapes) have been made
of certain sessions not covered by these Proceedings,
and for which permission was obtained from participating panel members. (Included are, for example, "ExtraSensory Perception and Man-Machine Communications," "Implications of Information Storage and Retrieval in the Human Brain," etc.) A copy may be obtained (for a fee) upon request.
We hope that the combination of advanced technical
sessions, tutorial sessions, discuss-only sessions, and
panel sessions provided a good balance meeting the
needs of all attendees active in hardware, software, and
management applications. However, as is the case with
every conference treating a field as broad as information processing, the 1965 FICC Technical Program emphasized some areas at the expense of others. We
endeavored to direct the emphasis toward topics of
current interest and ferment. For example, the increased
speed of components has intensified the imbalance between terminal equipment, memories and central processing units with memories forced to bridge the gap.
Thus, computer memory technology is emphasized in
these Proceedings to direct attention on problems and
prospects in this area. The topic of time sharing is also
one of current vitality. The two volumes contain several
papers related to this area, covering problems of hardware, software and usage. As information processing
systems become automated throughout the economy,
more attention is required by the problems of applications and management. This, too, is emphasized in these
Proceedings.
The vitality and success of the Technical Program
reflect the excellent efforts by conference participants,
attendees, and planners. We are grateful to all the
authors of the papers who made these two volumes possible. Indebtedness is also expressed to the chairmen of
the various sessions, including panel discussions, evening
discussion sessions, workshops and epilogue session for
the excellent results. So many did an outstanding job
that it would be unfair to try to single out a few who
did particularly outstanding jobs.

The structure of the Technical Program for the 1965
FICC emanated from a conference theme that emphasized the role of computers in modem society. The
technical program aimed at satisfying two goals: (a)
serve the needs of the entire membership of AFIPS
which has constantly expanded to include more diversified disciplines, and (b) focus attention on the role
played and the contributions made by computers to
American life and its institutions (cultural, economic,
etc.).
To implement the former goal, Friday sessions were
instituted. No record of these sessions has been kept in
either Part 1 or 2 of the Conference Proceedings. The
responsibility for organizing the Friday sessions was
delegated to member societies, under the supervision
and initiative of the Technical Program Committee.
This permitted coverage of specialized subjects for
which no time slots could be allocated within the main
three day technical program planned for much wider
membership participation. Sessions were also added to
serve the needs of management and marketing oriented
membership of AFIPS. These included broad-coverage
sessions (tutorial, panel and informal discussions)
spread throughout the conference. Records of two such
panel sessions-"The Computer Industry In The Buyer's Market" and "The Overseas Computer Market"are contained in this volume.
In an attempt to implement the latter goal, a Thursday session involving nine "keynote" speakers was introduced to focus attention on a wide gamut of disciplines
in which computers play an important role. During the
first two days the sessions were designed to cover the
more notable achievements of the past, and present,
and future trends. These were aimed at answering the
questions "Where are we now?" and "Where are we
going?". These sessions were then brought into focus
in the Thursday session to answer the question "How
are we going to affect the world around us?". The addresses by the nine "keynote" speakers are contained in
this volume. These papers dominate the second volume
of the Conference Proceedings, hence the title adopted
for it: "Computers and Their Impact on Society."
Part 1 contains formal papers delivered during the
first two days. This part also contains additional material presented the first two days. Besides the two panel
discussion records mentioned earlier, there is a record
of panel discussions on "The Future Of Electro Mechanical Mass Storage" and "Promising Avenues For Computer Research." It contains late papers which do not
appear in Part 1, and final versions of papers for which
the preliminary draft was erroneously printed. It also
contains an errata for Volume 1 which readers should
carefully note.
A novel experiment at the 1965 FICC was the Discuss-Only sessions where preprints of the papers were
vii

cuss-Only sessions on schedule. We are indeed also
proud of the paper reviewers who applied themselves so
diligently to assure high standards for the papers contained in these two volumes.
The assistance extended so generously by Gordon
Mitchell (NSA) , Morton H. Lewin (RCA), Harold
Judge (Edgerton, Germeshausen, and Grier), and John
McCrummen (Bunker-Ramo Corporation) is gratefully
acknowledged.
.
S. Nissim, The Bunker Ramo Corporation
T. B. Steel, Jr., System Development Corporation
Co-Chairmen, Technical Program Committee

Sincere appreciation is due to our collegues in the
Conference Committee, in particular the Technical Program Committee, for all the careful planning and hard
work that went into structuring the technical program:
Robert Gary (Secretary), Ben Ferber (Director, Management & Applications), L .D. Yarbrough (Director,
Software), J. R. Bennett (Director, Hardware), D. A.
Meier (Director, Member Societies).
The dedicated efforts of Messrs. Gordon Mitchell,
Wayne Cotten, Robert V. Davies, Robert S. Kiel,
J ames Mihalik, John F. Piontek, and David L. Yetter
at NSA made possible publication of preprints for Dis-

viii

FJCC 1965
Conference Committees
GENERAL CHAIRMAN
VICE CHAIRMAN
ADMINISTRATIVE COORDINATOR
TECHNICAL COORDINATOR
EXHIBITS

PUBLIC RELATIONS

SECRETARY-TREASURER
TECHNICAL PROGRAM

LOCAL ARRANGEMENTS

REGISTRATION

PRINTING AND MAILING
EDUCATIONAL ACTIVITIES
SPECIAL ACTIVITIES

WOMEN'S ACTIVITIES

FINANCIAL ADVISER
EXHIBIT MANAGEMENT

Robert W. Rector
Linder C. Hobbs
Marvin Howard
Sei Shohara
Richard B. Blue, Sr., Coordinator
Samuel F. Needham
Emmett R. Quady
William D. Orr, Coordinator
Richard Condon
Al Erickson
James Kehoe
Dale Lewis
Irwin Schorr
J. Bradley Stroup
Dawn Walker
Mark G. Singleton
Samuel Nissim, Co-Chairman
Thomas B. Steel, Jr., Co-Chairman
Robert K. Gray
J. Russell Bennett
Ben Ferber
John R. McCrummen
Donal A. Meier
Howard L. Parks
Lynn D. Yarbrough
Jerry Koory, Chairman
Al Bongarzone
Al Deutsch
Angeline Jacobs
Jay KleinBard
Frank Meek
Frank O'Neill
Arthur Rosenberg
Al Rosenthal
Eugene S. Gordon, Chairman
Weldon C. Dennis
George L. Jones
Robert J. McGill
J. E. McAteer
Marjorie. F. Hill, Chairman
Robert L. Albrecht
Gloria M. Silvern, Chairman
Carl N. Brooks
George F. Forbes
Irene E. Matthews
Shirley L. Marks, Chairman
Doris Uncapher
Fred J. Gruenberger
Ted Gatto
H. G. Asmus

Informatics, Inc
Hobbs Associates, Inc.
Informatics, Inc.
Scientific Data Systems
TRW Systems
TRW Systems
UNIVAC Division of Sperry Rand
SF Associates
Fox-Mathis
Librascope Group, General Precision, Inc.
IBM Corporation
CEIR, Inc.
System Development Corporation
General Electric Co., Computer
Department
Electronic Memories, Inc.
North American Aviation, Inc.
The Bunker-Ramo Corporation
System Development Corporation
System Development Corporation
Burroughs Corporation
Consultant
The Bunker-Ramo Corporation
National Cash Register
The Bunker-Ramo Corporation
Harvard Computation Center
Planning Research Corporation
Computer Sciences Corporation
Associated A ero Science Laboratories,
Inc.
American Institute for Research
North American Aviation, Inc.
Control Data Corporation
Stardust Hotel
Scientific Data Systems
The RAND Corporation
System Development Corporation
System Development Corporation
United States Air Force
System Development Corporation
Hughes Aircraft Company
Control Data Corporation
Consultant
North American Aviation, Inc.
Norair Division, Northrop Corporation
Litton Industries
TRW Systems
The RAND Corporation
The RAND Corporation
The RAND Corporation
AFIPS

THE AMERICAN FEDERATION OF INFORMATION
PROCESSING SOCIETIES
takes pride in presenting the

1965 HARRY GOODE
MEMORIAL AWARD
jointly to

George Robert Stibitz
for his contributions to, and pioneering efforts in,

• automatic computing,
• for independently proposing the use of the
binary system, floating point arithmetic,
memory indexing, and operation from a
remote console, and
• for designing the first operating programcontrolled computer.
and

Konrad Zuse
for his contributions to, and pioneering efforts in,

• automatic computing,
• for independently proposing the use of the
binary system and floating point arithmetic,
and
• for designing the first program-controlled
computer in Germany-one of the earliest
in the world.
Las Vegas, Nevada, December 1, 1965

Chairman, AFIPS Board of Governors: EDWIN L. HARDER
Chairman, AFIPS Awards Committee: SAMUEL LEVINE
Harry Goode Memorial Award Committee: JERRE D. NOE, Chairman, SAMUEL N. ALEXANDER
ALSTON HOUSEOLDER, JON C. MCPERSON, WILLIS WARE

Biography-George R. Stibitz

ON SEPTEMBER 11, 1940, George R. Stibitz dem-

features of modern computers such as binary arithmetic,
indexing, self-checking, and floating point arithmetic.
Among the "Stibitz computers," as they are sometimes
called, are the Relay Ballistic Computer for NDRC
( 1943/44), whose original model now is at Fort Bliss,
Texas, and large relay computers now at NASA's Langley Research Center and the Army's Aberdeen Proving
Ground.
George Stibitz' activities have, over the years, encompassed a multitude of scientific subjects, from
propeller vibration computations and theory of errors
in counting computers, to guided missile simulators,
information theory, and learning machines. His investigations have found expression in many published
works, in unpublished reports and memoranda, and in
talks and lectures before professional societies and
institutes of higher learning. Among the earliest of his
many patents, George Stibitz' patent for a Binary
Computer dates back to 1943.
At present, he is associated with the Department of
Physiology, Dartmouth Medical School, where he is
principally concerned with applications of mathematics
and computers to biomedical areas and computer program models of the nephron and cilia.

onstrated for the first time the remote operation of
a relay-type computer from a console at Dartmouth
College in Hanover, New Hampshire, over telegraph
lines to the' main frame in New York. This Model I
Relay Computer climaxed George Stibitz' intensive
investigation into computer principles, theory, and
existing technology. The first model was followed by
five succeedingly more advanced relay computers,
built during and after the Second World War.
George R. Stibitz received a Ph.B. degree from
Denison University in 1926, a M.S. degree from Union
College in 1927 and a Ph.D. in 1930 from Cornell
University. From 1930 to 1941 he was a research
mathematician at Bell Telephone Laboratories; from
1941 to 1945, a technical aide with the National Defense Research Council of the Office of Scientific
Research and Development. Since then, George Stibitz
has rendered valuable services as a consultant to many
clients in a wide variety of scientific endeavors.
George Stibitz virtually designed the first two relay
computers and made major contributions to later
machines of this series, two of which had most of the

xii

ACCEPTANCE SPEECH OF
GEORGE R. STIBITZ

I need not tell you how greatly I appreciate the

first of the binary computers. Here are a couple of comments from news releases about a remote control
demonstration pre-dataphone, from Hanover, New
Hampshire, to New York City in September 1940.

honor of your recognition, and I thank you and the
Award Committee for granting me the Harry Goode
Award.
Naturally, since the award has been made for pioneer
work in the later 1930's and early 1940's, my thoughts
turn to that period, and I have looked back over a few
records of those days.
I recall the first binary adder I know of. It was
built on our kitchen table in 1937. The arithmetic unit
consisted of two telephone relays. The storage-with
a capacity of two bits-was two more. The input mechanism was a pair of keys made of sheet metal and
the output was presented visually by two flashlight
bulbs.
From this model to a binary computer for complex
arithmetic was a fairly rapid step, and in a few months
I had designs for such a computer, using the excess-3
code, as well as plans for a more general algebraic tapeand-keyboard controlled computer.
The Bell Telephone Laboratories had need for the
complex computer and it was built and put to use in
1939, but the astronomical cost~cIose to $20,000-was
judged excessive for a computer, and the more elaborate
plans were not used until the war.
I am thinking, too, of the contributions of Sam
Williams and of Andy Andrews to the early developments, especially to the series of relay computers
built dunng the war years.
I t was interesting to glance back at some of the
opinions of experts and of lay people in regard to the

In the New York Sun, September 9, 1940, I find:
The Bell Laboratories "have not the remotest idea that
with their machine they ever will be able to perform
a service for any outside industry and they do not
foresee the time when Johnny, aged 12, will be able to
pick up his telephone receiver and ask the operator for
the answer to seven times nine."
But I find a more optimistic view of the same demonstration expressed by a reporter who signed himself
"L.W." in the Newark (New Jersey) News, September 9, 1940. I quote: "And if this may seem a restricted
use, just remember there was a time when few people
thought the world would have much use for flying
machines. "
Contrasting the views of these. two sources, I feel
there isa moral of some kind to be drawn, but can't
quite make up my mind what.
In my present occupation of applying mathematics
to medical and physiological problems at the Dartmouth Medical School, I have had constant contact
with the computer there, and stand in awe of the accomplishments of you people and your colleagues in
matters of speed, capacity, and convenience of the
COITlputers now available.
Again, I thank you and join in remembrance of Harry
Goode in whose name the award is given.

xiii

Biography-Konrad Zuse
built between 1939 and 1941.
Following these achievements, Konrad Zuse constructed two special-purpose models of this computer:
the Model S-1 for aircraft-wing measurements, and the
Model S-2. This latter model can be called the first
process control computer because the results of about
100 measurement points were continuously sampled
and transmitted to the machine .. During this time he also
began work on the Model Z-4, the only computer he
was able to save; all of his other computers were destroyed during the Second World War.
Undaunted, he continued his work on relay computers which, in later years, gave way to vacuum-tube
and solid-state machines. As early as 1945/46 he developed the "plan calculus," a forerunner of modern
programming languages.
Starting his own company in what is now West Germany, the Konrad Zuse Kommanditgesellschaft, in
1949, he became an industrial pioneer, with computers
gaining in importance in the German market. While he
published relatively little, his thinking strongly influenced the development of the computing sciences in
German-speaking countries.

KONRAD ZUSE is the German pioneer in automatic computing. A building engineer by training, he
soon found the numerous computations for statics a
tiring and laborious undertaking, which finally led to
his decision to construct a mechanical aid. His early
work on these computing "mechanisms" covered the
period from 1934 to 1936, during which time he laid
the foundations for the Models Z-1 and Z-2 Relay
Computers built between 1936 and 1940. In 1941 he
sought a patent for his "Rechenvorrichtung"; the patent was finally granted in 1952.
A native of Berlin, Germany, Konrad Zuse received
his technical education in his home town at the Technische Hochschule (Berlin University of Technology),
from which he later received an honorary Doctor of
Engineering degree (Dr. Ing honoris causa) in 1957.
Dr. Zuse also holds several other important patents in
the field of computing science., Independently of others,
he proposed the binary system and the floating decimal
point concepts as applied to computer design and built
the first program-controlled computer in Germany,
the Model Z-3. This machine had 2,600 relays, and a
64-word memory of 22-bit words. The computer was

xiv

ACCEPTANCE SPEECH OF
KONRAD ZUSE

was very happy to receive the invitation of the
American Federation of Information Processing Societies, to come to Las Vegas, first because of the
honor attached to this invitation, but also because this
provided me with the opportunity to come to Las Vegas
without my wife being able to object.
I regard it as an exceptional honor to receive the
Harry Goode Memorial Award. I do not only consider
this as a personal distinction, but I am also highly
gratified to see that an American organization should
thus appreciate the achievements of European pioneers.
For these reasons, I think it appropriate to tell you
some experiences about the history of European computer development.
Let's start with some general considerations on the
calculating machines. In contradiction with the general
opinion that Pascal and Leibniz were the first ones
to have developed calculating devices, it was found by
recent historical research that the German Schickard
was the forerunner for some 30 years.
Most of you will have heard the name of Babbage.
He is the real father of the program-controlled computers. More than a· century ago he had made the plans
and started the construction of a digital computer.
Simply because the technical means available at that
time were not adequate, he failed to achieve his goal.
Only a few people realize that he already knew
the conditional order.
As forerunners of the present computer development,
some other Europeans can be mentioned. For instance,
the English mathematician Turing who, from the purely
mathematical logical point of view, used the concept of
a universal calculating machine. He used this concept,
now known as the Turing machine, when analyzing
the computability of mathematical functions.
The Frenchman Valtat had already back in 1936
applied for patents on the idea of binary calculating
machines. Another Frenchman named Coufignal had
developed in his doctor's thesis, only recently publicly
known, certain ideas concerning the program control,
logical operations and binary system for application in
computing machines.
Specific development of computers as we under-

stand them now took place between 1935 and 1945.
Quite independently the work proceeded both in the
United States and in Germany. Very remarkably, both
approaches were not made by specialists in the field
of calculating machines, but by outsiders. Since both
developments were carried out practically during the
same period, it may be useless to try to determine which
machine was the "first" one.
I am convinced that you are better informed about
the development in this country than I could be, and
therefore I would like to give some more information
about the development going on in Germany until 1945.
In the early thirties, I was a construction engineering
student at the Technical University of Berlin. You will
probably know that, in this field, extensive calculations
have to be carried out, especially for indeterminate
systems. This led me to think about the possibility of
designing computers. My goal was to be able to carry
through, fully automatically, complete calculation sequences. I did not know anything about computers,
nor had I ever heard about the early work of Charles
Babbage. Thus-unprejudiced-I could go new ways.
In order to illustrate the opinion of the manufacturers
of calculating machines at that time, I would like to
mention a telephone conversation which I had in 1937
with one of those manufacturers. He told me that it
was indeed wonderful that I as a young man had dedicated some time and efforts to the development of
new ideas, and that he wished me all the best for
possible other inventions, but stated that in the techniques of calculating machines all feasible solutions
were already exhausted. Therefore, it would be absolutely hopeless to come up with any new ideas. In
addition, he asked me whether my machine was based
on the "sequential addition principle" or on the "one
times one table." To this I replied that for my machine
this was of no importance whatsoever. Here you should
know that at that time the specialists of calculating machines were divided in two schools of thought, each
applying either principle. According to the opinion
prevailing at that time, only a lunatic could make a
statement that this difference was irrelevant for his
design. Nevertheless, the manufacturer mentioned

came to my workshop and I was finally able to convince him that, in a machine operating on the binary
principle, this was indeed irrelevant.
Due to the lack of interest encountered I started on
a private basis, supported financially by a few personal
friends on a very limited scale. Later I received some
backing from research institutes and from the German
aircraft industry. To begin with, I built various allmechanical models. However, I soon realized the limits
of mechanical design and, therefore, adopted the electromechanical concept. This led eventually to the sequencecontrolled relay computer Z3, which was completed in
1941.
The main characteristics of this machine are: sequence control, binary system, floating-point and
electromagnetic relay technique with wide application
of switching theory.
This machine was complete and ready to operate.
Mainly it was used for test computations. The programs served to determine the eigenvalues of complex
matrices; this was a problem of particular interest to
certain aerodynamics specialists.
Apart from this one, another special purpose prototype was built, which had a fixed program store wired
into rotary switches. It served the task of evaluating
wing measurements of a missile assembly line, and
operated 16 out of 24 hours for 2 years. The problem
was to multiply about 100 gauge values obtained on
missiles with certain characteristic numbers in accordance with an aerodynamical program, in order to get
adjustment values for the positioning of the wings.
In a more advanced version, the gauges were read
automatically via rotary switches, which transferred
their positions into the computer. This device may well
have been the first computer application for process
control.
Proceeding from the universal computer Z3, an improved model Z4 was built, which was essentially completed and could operate on simple programs, when
the war came to an end. Alone among the whole series,
this machine, Z4, could be rescued. After some enlargement, it was later installed at the Swiss Technical University in Zurich. The historical model Z3, which had
been destroyed during the war, was later reconstructed.
It is now in the German Museum of Science and
Technology in Munich.
Due to the situation in postwar Germany, any further
work on the subject was rather difficult at first. Also,
the task of designing new computers became teamwork.
As to myself, I am afraid I was largely occupied by
other problems due to my function of building up
and running the Zuse Company. I could, however, hold
some essential personal share in the development of a
general-purpose automatic digital drawing table, which
went into production in 1959, together with other spe-

cial devices, for geodeticspurposes.
I would like to mention the fact that, in 1938, my
friend, Dr. Schreyer, set himself the task of redesigning
my relay computer by means of electronic components.
He developed a set of basic circuitry for solving the
fundamental logical operations and built a prototype
binary arithmetic unit for 10 digits. During the war,
we submitted the design drafts for an electronic computer with 2000 tubes to German research authorities,
but their reaction was negative.
I would now like to take the opportunity to cite the
name of Dr. Dircks. He had, at a very early stage,
produced essential ideas in the field of magnetic storage devices. During World War II, he constructed a
prototype of a small magnetic disc file.
Apart from this work on hardware, I was able to
devote some of my time to software philosophy as
well. The relay technique induced me to do some work
in the field of switching theory. From this I gained the
fundamental knowledge that any information may be
dissolved into true-false values .(today called bits). I
introduced the general concept of computing as follows:
To compute means to derive new information from
given information, according to a prescription or better
said algorithm.
At the outset, I developed some drafts for logical
computers; for one of these we built a simple prototype.
From this resulted the necessity to develop first a
general formula language which might be used to formulate even the most general and complicated computing sequences. This work was finished in 1945 with
the creation of the so-called Plankalkill (Programming
Calculus). Essential features of this were:
Consequent adherence to a systematic representation of all information structures.
Extension of the function concept.
Introduction of the sign "results in" (=».
Incorporation of logical operations and functions
(calculus of propositions, calculus of relations
etc.).
In the hardware which I developed until 1945, I
strictly avoided the use of conditional orders. The
theoretical analysis made proved to me that a .single
wire connection feeding back from the arithmetical
unit to the program controlled unit would enlarge the
logical capabilities of the machine to an extent, the
consequences of which were difficult to foresee. I
elaborated the various possibilities of such a feedback
on a pure theoretical basis and introduced them in the
software philosophy. While doing so, I clearly found
that if once the step is made towards the conditional
order, the door is open for consequent development
leading to the present-day and future computers which
finally, step by step, take over large tasks of the human
xvi

brain. However, I did hesitate to be the one who made
this dangerous feedback connection at first.
The Programming Calculus served partly as a starting-point for our modern formula languages, such as
Algol and others. Its basic ideas may not, however,
have unveiled their full implications until now. As you
will agree, we have at present arrived at a point in time
wher~ the more specialized formula languages in use
today have reached some limits. The hardware which
I developed in those days has now become of historical
value only. On the other hand, I tend to assume that
the software ideas which make up the Plankalkiil, or
Programming Calculus, may be of some importance in
our day. I cherish the hope that they will still bear
some fruit.
I also devoted some-perhaps interesting-thoughts
to the self-reproducing machine. In the United States,
this problem has been treated mainly from a mathematical point of view. My own thinking was independent
of this and moved more in the field of engineering. It
led to the concept of what may be called a "technical
germ cell." Again I hope that some day it will gain a
certain actuality.
The brief report that I could give you includes mainly
the period until 1945.
You all know that after World War II, an up growing
development took place in the field of digital computers. In Europe this postwar development was ini-

xvii

tiated mainly in England. However, I don't feel competent to select from the large amount of work done
the most important approaches.
Also in Germany some development was started by
various groups. Before the computer industry gained
some importance, there were research groups at universities and other institutions who took the initiative.
Amongst those who participated in this work, I would
like to mention Dr. Billing, who designed at a very
early stage a drum memory.
Besides the work done in England and Germany,
there were some other computer developments carried
out by the following European countries: Austria, Belgium, France, Netherlands, Sweden, and Switzerland.
These are the ones known to me, and it is possible that
other countries unknown to me have also carried out
some original development work.
We Europeans have a considerable admiration for
the research done in this field in the United States, as
well as for the American computer industry.
On the other hand, 1 am glad to see your interest in
our European approach and I am very grateful that I
was given this opportunity to talk to you on this subject.
I do sincerely hope that international cooperation will
continue to grow and that in this activity sector ideas
coming out of good old Europe may still contribute
to our common progress.

The mighty man-computer team
IDA RHODES
National Bureau of Standards
Washington, D.C.

that should another ~1erlin arise, capable of constructing a lifeless contraption for simulating human thought,
he would run out of all the material available in the
Cosmos, long before he succeeds in completing his very
first specimen.
The elation, which some of us may feel at such
largesse on the part of Providence, rapidly subsides as
we reflect on how few members of our species either
fully appreciate, or strive to make use of, their prodigious birthright. Regard the humble amoeba. With its
single tiny cell, it manages to find its proper environment; to gather, ingest, and digest suitable food; to
. eliminate its wastes; to grow, to mature, and to produce
young. How many of the earth's three billion human
denizens care to utilize their stupendous mental powers
for higher or nobler aims than the unicellular amoeba?
We trumpet with loud fanfare the blessings of our
huge array of devices, invented for the purpose of
extending and enhancing our natural prowess, as well as
of freeing us from back-breaking, time-consuming tasks.
Let us bear in mind, on the other hand, that every tool
is a two-edged sword. If its wielder be animated by
vicious motives, it can turn-in his hands-into an
accursed weapon of wanton destruction. Impressed by
man's mighty intellect and the horse's amazing strength,
our ancestors were inspired to blend the two into the
image of a devastating Centaur. It is now within our
power to replace this violent monster by an incomparably more efficacious, yet supremely beneficent,
Megataur. Embodying the awsome potentialities of the
high-speed computer and guided, at all times, by the

If the Martians, Venusians, or Plutonians ever bother
to observe the antics of their neighbors on the Planet
Earth, they must be vastly amused by our attitude
toward the digital automatic computer, the DAC for
short. While professing to be completely baffled by the
frenzied rush on the part of the lemmings to be drowned
in the sea, we-earthlings-have perversely chosen to
denigrate the unsurpassable human brain by affixing
to a pile of wires and tubes that ludicrous title "The
Thinking Machine." What a churlish way to thank
Mother Nature for our bountiful mental endowment
which dwarfs into insignificance those fabulous gifts of
magic, lavished upon their favorites by the doting fairy
godmothers.
Let us take inventory of our divine heritage. The
human cranium possesses a storage capacity equivalent
to at least a thousand billion binary digits. Supplementing this opulent installation, there are five known input devices-our magnificent sense perceptions-and
perhaps additional senses, of whose existence we are
only dimly aware. Maintained by a superbly efficient
system of physical organs, our mind is enabled to exercise an astounding number of sublime functions. The
myriads of impressions, which we receive through our
senses, are being constantly shuffled and miraculously
combined into concepts, whose count exceeds by far the
total number of elementary particles in the entire universe. Yet, this colossal aggregate is actually infinitesimal in comparison to the fantastic number of ideas
which our brain can engender by continuously associating various concepts. We are thus led to the conclusion

The 'Mighty Man-Computer Team

highest dictates of an exalted human conscience, the
new image could become a source of radiant hope for
our strife-ridden, despair-laden world.
It is gratifying to learn that a group of high-minded
medics have united to form "Physicians for Social Responsibility." It is easy to predict, that the members of
our profession, who were deeply stirred by the eloquence of John F. Kennedy's inaugural address, or
by the loftiness of Pope Paul's plea for peace, or by
the earnestness of President Johnson's appeal to join
the Great Society, would be eager to endorse the "Megataur for Social Responsibility."
We might start by borrowing a custom of the ancient
Hebrew scribes, who underwent an elaborate daily
ritual of self-purification, before assuming their sacred
task of copying the Scriptures. Before outlining what
constitutes, from my point of view, a similar ritual for
Automators, I would want to make sure that all visitors'
guns were checked at the entrance.
We are told that when the great de Forest-regarded
as the father of television-observed the fare being
dished out for the benefit of the captive viewers, he
exclaimed in despair: "Heavens, what have they done
to my child?" I suspect that the inventors of DAC are
moved to utter a similar cry of distress, when witnessing
the daily abuse and misuse of their illustrious brainchild.
It can hardly be a source of pleasure for them to be
told that a prospective user of their wondrous prodigy
requires only a brief exposure to mumbo jumbo in
order to match the performance of an Isaac Stern on
a Stradivarius. We are put in mind of the visitor to a
European Museum. Glimpsing the piano of the composer Liszt, he decided to display his own virtuosity
as a pianist; whereupon he voiced his conjecture that
previous visitors must have been inspired to similar
action. "I would not say so, sir," remarked the attendant. "The last person who came through here was
Paderewski. He claimed he was not worthy of touching
the keys of this instrument." Now I submit, ladies and
gentlemen, that in order to become an accomplished
automator, one must go through as much preparation
and training, as-say-a skilled physician, or competent lawyer. Members of these and numerous other
crafts have wisely established rigid criteria for licensing
their applicants. From my point of view, it will be
a red letter day in the annals of electronic computation,
when means will be found to keep the rank amateurs
away from our multimillion-dollar DACs. Those perennial bunglers tie up the machines with their error-laden
routines which soon assume the visage of a ferocious
hydra, since each time they undertake to correct one
of their goofs, they manage to introduce seven brand

new ones.
My next suggestion for a sacrificial offering will raise
quite a few hackles. But remember, no shooting! The
exceedingly simple and uniquely suitable language with
which each inventor endows his DAC is calculated to
secure maximum economy and expediency of performance. Yet it is consistently being discarded in favor of
a spate of grotesquely time-wasting jargons. Disregarding the fact that superbly qualified teams of mathematical geniuses have not yet succeeded-after millenia of
intensive effort-in creating an unambiguous universally
accepted terminology for their science, the high priests
of mumbo jumbo insist that their quest for artificial
languages must go on, because of the great diversity
of machine-code structures and the massive infiltration of the aforementioned breeders of hydras. I shall
not take the time today to refute the fallacy of their
contentions; instead, I should like to sponsor the alternate proposal of standardizing the machine codes.
When Artur Rubinstein is invited to perform in Tokyo
or in Moscow, it is not necessary for him to tote his
Steinway to those distant places. Musicians have
learned long ago the wisdom of regulating the basic
notes of the scale, as well as the forms of the instruments that produce them. Although we have not yet
reached the stage where our fraternity can undertake
to compile a roster of instructions that will endure for
all time, we have amassed enough experience and authority to agree on an excellent list which would serve
quite well for at least a decade. I venture to predict
that our demand for a universal adoption of this set
would be greeted with a sigh of relief by all makers of
DAC, since they would be liberated, at last, from the
murderous treadmill of software construction and would
be able to invest the considerable amount of money
thus saved on truly effective data handling attachments.
If I still have any friends left in this audience, I hope
they will bear with me, as I mention still another
method of self-purification. The Russian people have
an apt saying: "The house is still in the process of
completion, and the rightful owners have not yet moved
in, but the cockroaches are already in full possession
of the premises." Our vital and honorable profession
is in grave danger of being invaded by a horde of selfordained geniuses who, under the guise of earnest
seekers of scientific truth, may succeed in foisting their
spurious wares upon generous but gullible sponsors.
As an example of a field, highly susceptible to this sort
of infestation, I might mention Information Storage and
Retrieval, including Machine Translation. The only
hope of attaining even a modicum of success in this
and similar endeavors lies in the complete and cordial
cooperation of all interested parties, not merely on the

The Mighty Man-Computer Team
national, but also on the international, level. No isolated,
uncoordinated group of workers can ever come up with
an acceptable answer to the heartbreakingly complex
problems involved in those fields, just as no set of
isolated, local meteorological observations could possibly yield reliable long-range weather forecasts. Let
us make quite sure, therefore, that the ironic tale of
the "The Emperor's New Clothes" is not reenacted
right in our midst, lest we acquire the unsavory reputation of unscrupulous schemers and heartless mulcters
of the public till.
The precious machine time now being scandalously
wasted on the sprawling monsters of the milksop coders;
on the gibberishes, broadcast from the Tower of
Babel; on the devious machinations of the Emperor's
tailors-might be almost sufficient to compute the
location of that elusive fulcrum sought by Archimedes
in order to move the world from its orbit.
As soon as we shall have succeeded in putting our
own house in order, we can start setting up the Megataur
for the vital business of cleaning out the Augean stables
of human misery. Each generation is prone to put the
blame for the hideous state of those stables upon the
bigotry, hatred, and greed of its predecessors and to
insist that an army of Herculeses would not suffice to
cope with the accumulated filth. In fairness to our forebears, it should be pointed out that they did not possess
DACs to help them solve the staggering social and
economic problems which have been plaguing humanity
for countless centuries. Lacking such equipment, even
their noblest efforts had to resemble-perforce-the
awkward performance of a tyro mathematician as,
pencil in hand, he struggles to calculate a solution by
the method of relaxation. Unable to assess properly the
impact of all the conditions existing in the pertinent
domain, he pounces upon some troublesome spot and
labors assiduously to relieve the tension in its immediate neighborhood. Unfortunately, he only succeeds in
raising a far greater protuberance in another, hitherto
calm, region. He rushes to the new center of disturbance to apply the same panaceas, and this time causes
a huge geyser to erupt in still another portion of the
domain. The purity of his intentions is unquestionable,
the diligence of his efforts is admirable, but the quality
of his results is deplorable. If we now expand our vision
to embrace the entire gamut of human activity and
watch, in dismay, the vast mass of mankind, writhing
and seething under the lash of unremittent hunger, pain,
and indignity, we realize that any attempt to calm down
permanently this restless agitation is quite futile without
the aid of titanic implements that can help us analyze
and remove its underlying causes.

The present models of the sadly miscalled "Giant
Electronic Brain," though quite adept at smoothing
out the bumpy domain which had confronted our tyro
mathematician, is pitifully inadequate for tackling the
formidable complex of matrices representing the all but
hopeless tangle of human affairs. However, anticipating
the advent of its far more overpowering successors, we
may now dare to aspire to the solution of some of the
most pressing problems facing our fellowmen. For the
present, we can rely only upon the penetrating force of
an exalted dream. I shall not presume to spell out for
this audience what the nature of that dream ought to
be. In the words of President Kennedy: "For this, every
man must search his own soul." I would like, however,
to relay an anecdote about Abraham Lincoln, which
throws an interesting light upon the topic of self-interest.
While riding with a friend along a country lane,
Lincoln was engaged in a discussion relating to the
various motivations that underlie human conduct. The
friend maintained that only altruism may be regarded
as an acceptable guide for man's behavior, whereas
Lincoln argued that the dominant impulse necessarily
resides in selfishness. Suddenly the air was rent by the
piercing squeal of a pig trying to extricate its head from
the slats of a farmer's fence. As Lincoln jumped out
of the buggy and ran to deliver the animal from its
plight, his friend taunted him good-naturedly: "I suppose you are going to tell me that freeing this pig was
an act of pure selfishness on your part." "It most certainly was" asserted Lincoln. "If I had failed to come
to the aid of that poor creature, I would not have been
able to sleep all night."
I think all of us will agree that Lincoln's interpretation of selfishness may serve as an excellent definition
for "social responsibility." Over a period of countless
generations, the term "reaching for the moon" was used
as a synonym for an absolutely unattainable ideal. Yet
recent events have deprived it of such connotation. One
need not be a Walter Mitty to imagine the dizzying
heights which may be reached by humanity, if each of
us resolves to adhere to the principles of social responsibility at all times and under all circumstances. Such
unswerving dedication is bound to instill in our subconscious minds an abiding habit that would automatically impel us to seize upon every opportunity to
elevate the stature of the Megataur.
Let us take for our motto, then, a paraphrase of the
immortal words, uttered by our beloved leader, martyred two years ago: "Ask not what the machine can
do for you. Ask what you, in conjunction with the
machine, can do for your country and for the world."

The computer and our changing society
SIMON RAMO

Vice Chairman of the Board
TRW Incorporated
Redondo Beach, California

Those responsible for this convention are to be
congratulated for including a whole day's session on
the impact of computers on society. Usually when engineers and scientists meet the subject matter is almost
wholly technical. In a way this is strange because the
common, quick definition of engineering is "the application of science to society's needs." If this definition were really to be taken seriously, it would mean
that those who practice engineering would seek to be
equally expert in science and in society, 'Since one can
hardly be professional about applying science to something he does not understand.
Engineers are only occasionally expert in the problems, needs, and organization of society. Conversely,
knowing very little about science does not automatically make someone outside the engineering profession
any more competent in social problems than the engineer. This is a great shortcoming of our world as
we head for a much more technological civilization.
There is, as a matter of fact, a missing professionthe "socio-technologist."
All in all, we may not be as prepared as we would
like, but it is particularly urgent at this time to discuss
the impact of the computer on our lives. More broadly,
we should consider the impact of the new technology
which involves all aspects of the handling of intellectual
and informational tasks by electronics, and for which
we will often use· the word ."computer" as a short,
though inadequate, title. It is especially timely and
important. to take up the computer's relationship to
society, because the computer is rapidly replacing nu-

clear energy and space as the leading technological
item of confusion and fear in the public mind. We cannot today picture with completeness and precision the
entire future effect of computers on our civilization.
However, we can endeavor to eliminate some common
misconceptions which are on their way to becoming
well-established, harmful myths.
WILL THE COMPUTER REPLACE MAN?
This may be the greatest fear. It is apparent from
the frequent appearance in the public press of statements such as "A comp\lter can't think" or "A computer isn't creative-it can only do what man directsit has no mind of its own!" These are defensive words
arising from the myth that the computer is a. competitor to man. It is apparently important to keep
reassuring ourselves that we are intellectually superior.
Thus, we rejoice when we hear that an automatic
device somewhere has failed. When the electric power
blackout took place in the N ortheastem United States
recently, many hoped to hear, and were anxious to
spread the word, that a computer, to which we presumably entrusted too much responsibility, had goofed.
"Please, God, don't let it 'turn out to be the result of
a human error," was the prayer many intoned.
But that man and the computer are competitors is
a misconception. Electronic information handling technology is presenting new ways to us to acquire, store,
process, disseminate, and utilize the information that
makes the world go round. It is making possible improved systems of production, banking, transportation, and education. The true concept is change5

The Computer and our Changing Society

change simultaneously offering tw,o things: potential
benefits and potential dislocations. A mature society
will work at minimizing the negative reaction so as
to emphasize the benefits. Properly handled, the computer in replacing man represents a small effect, and
the potential benefits, a large effect. At any rate, fear
and misconception will certainly work against realizing
the benefits.
What the computer makes possible is not primarily
a replacement of man or competition with man, but
a new man-machine partnership. It enables an extension of man's intellectual and information handling
capabilities and hence enhances man. The combination, exceeding the capabilities of unaided man, can
better meet society's needs and can attain higher
achievements. As to man's specific occupations that
might be affected by his partnership, the result will be
a new and broader spectrum of jobs, new professions,
and greater satisfactions.
The X-ray machine did not compete with or replace
the physician. It broadened the practice of medicine.
It made it a better profession, a more intellectual one,
a more useful one to society. It brought a requirement
for technicians and apparatus and facilities not previously existing, all economically justified by the new
technology.
The book is a machine having to do with extending
man's intellect. When the concepts of printing and
books were born perhaps some considered books as
competitors to teachers. In a sense, they did replace
something that the human educator previously had to
accomplish alone, without books. But they. succeeded
in broadening the dissemination of knowledge, increasing communications of thoughts between people,
leading to a much broader and larger educational
system. Books made possible the employment of even
more teachers, who brought more benefits to society
than were economically achievable without the
machine-in this instance, the book.
ARE EDUCATIONAL REQUIREMENTS
FOR THE NEW TECHNOLOGICAL
SOCIETY UNATTAINABLE?
There is a growing alarm that the educational requirements for the decades ahead simply cannot be
met. The technological life toward which we are headed
requires increasingly broader education of every young
person to prepare him to make a living, and to be a
good citizen. A larger fraction of adults must now
continue education to keep up with the rapidly expanding order of change and remain productive. Meanwhile it is seemingly becoming more difficult to provide

this education. The cost of facilities and the demands
for educators are escalating, and shortages seem certain to grow. This educational crisis is often blamed
on technological advance. The technological society
generates, it is claimed, impossible demands for education and hence guarantees its own failure and collapse.
A breakthrough in educational concepts is clearly
needed-something comparable with printing and the
book. Advanced electronic technology offers precisely
the kind of new concepts needed to revolutionize the
educational. process. Human educators can be assisted
by networks of electronic· facilities that are backed by
an educational industry that does not even exist today.
Educational experts will be able to plan present, test,
and analyze with great enhancement of their informational and intellectual powers, rising to a new plateau
of accomplishment. Here the man-machine partnership
can provide a match between society's needs for all
kinds of education, and society's ability to supply
that need.
Embryo teaching machines of today are as far away
from the full use of technology in education as the first
stone tablets are from today's television network.
Matured computer and electronic networks, applied
under the skillful direction of educators and engineers
working together, can provide new forms of education in the home, in schools and in industry. Educational communication satellite systems can provide to
a special network in the homes of the nation a choice
of hundreds of different courses of study chosen
simply by pushing the right buttons on the home educational TV set. Carefully. presented programs available
when called for can involve student identification and
participation, the answering of questions by pushbutton, the monitoring of answers by "live" experts, and
a record of results.
In schools, material can be presented not only by
the human educator in person, but through audio-visual
devices which can automatically speed up, slow down,
or switch to a completely different presentation, all in
automatic response to the student's apparent ability to
follow the material. That is, the presentation can
periodically include questions to the student, the answers to which can influence further presentations to
suit the student's pace. Computer systems can keep
track of the progress of millions of students. These
systems can compare progress against estimates and
can make possible statistical analyses and a type of
creative planning not now remotely practical. At the
same time, an individual student can have synthesized
for him presentations or tests completely unique to his
particular. requirement.· These can be determined by the
virtually instantaneous availability of a full record of

The Computer and our Changing Society
progress and a comparison of that record against alternate courses for him to take in the future.
Electronics does not separate the student from the
human teacher any more than the electrocardiograph
keeps the physician from having direct contact with
and interest in his patient. On the contrary it enables
him to "listen" to his patient's heart with greater
skill. Comparably, the human educator will be able to
consider an enormously larger number of facts about
both student and course with greater accuracy and confidence. He will be able to make available to the student
more material with greater efficiency and with a much
broader selection. He will be able to reach the eyes,
the ears, and the minds of students, children and
adults alike, everywhere. He will be able to propose
new concepts in education and to check his plan to
discover how it worked, and then alter and improve
it.
In the coming technological society, education can
become the greatest occupation of man and his greatest
preoccupation as is required to meet the challenges of
the age. The educational profession will expand into a
larger number of specialties, and it will be equipped to
do more research to achieve important generalizations
to guide its memcers in that educational acceleration.
Educators will have the support of a huge industry that
provides systems engineering, communication networks, information dissemination, storage and retrieval, and analysis techniques capable of supporting
the higher educational plateau.
I t is a myth to think that rapidly advancing technology is creating an impossible dilemma in education.
It is rather that the crises that have been developing
for years in education may at last be attacked with
a scope equal to the task by the utilization of advanced
technology.
WILL THERE BE NO NEED
FOR UNSKILLED LABOR?
The leading "dark horse" candidate for the myth of
the century may be this: modern technological society
eliminates the need for unskilled labor. Since, no matter how technology is used for the good of society, not
everyone will be well educated and competent to work
at highly intellectual pursuits, this myth suggests we
will permanently have a growing number of unskilled
and unemployed citizens. I venture to predict that after
some initial dislocations, and within the period of a
decade, we shall commence to realize that, far from
there being a shortage of jobs for unskilled labor, there
may well be a shortage of unskilled labor to fill jobs.
To see why this is so let us move ahead to the year

2000. Let us also assume, in complete consistency with
the premise of the myth, that unskilled labor will no
longer be needed because we will have reached an essentially fully automated society. The factories will
turn out all of our material needs through nearly automatic operation with little intervention by man. There
will be moving sidewalks and automatic rapid transit
cars in the cities which, like today's automatic elevators, will perform without human operators. None
of us will use coins or currency anymore. We will instead assume that when we buy a necktie or a piece of
land we simply hold our finger against a little window
so our fingerprint can be scanned electronically, then
in the record-keeping facilities thousands of miles away
something will be taken off our account and put on
someone else's account. Let us assume that automobiles, missiles, and houses are designed by automatic
computer programs. The tests taken of each of us in
the hospitals will be automatically recorded and analyzed, and a treatment prescribed with little intervention by human operators or analysts.
This is an exaggeration, an extreme description of
a fully automated age. However, this route enables us
quickly to arrive at a series of important points. The
world fitting our description is reachable only by a
complete redoing, updating if you will, of our entire
national physical resources to take advantage of
technology to reach the fullest automaticity conceivable. To achieve such a system, even if it were possible
or desirable, would require an expenditure per year
far exceeding our gross national product. It would
mean virtually creating a whole new nation full of
new expensive facilities and resources and a newly developed and implemented way of operating them.
Our cities, factories, hospitals, schools, and transportation systems would all have to be rebuilt. We do not
have the total resources in manpower, skilled and unskilled, to accomplish this transition by the year 2000.
Even if it were economically or socially sound, even if
there were individual or cooperative incentives to seek
to achieve it, the total cost would be too great. We
simply don't have what it takes.
But this is another way of saying that, in a basic
sense, a major transition of this sort is economically
unsound. What really will happen will be much more
sensible. To use advancing technology to improve all
aspects of our society will involve the proper use of
people and things, men and machines. The only practical and reliable plan is to make an optimum selection
from these two categories for the tasks to be performed.
Now, man can be produced with relatively cheap labor,
and can be trained to do an enormous variety of tasks
with his brains and senses while having physical mobili-

The Computer and our Changing Society

ty, for a rather reasonable initial and annual maintenance cost. By comparison with a certain class and
variety of jobs man can do, a machine designed to do
the same tasks becomes absurdly expensive. We can
abbreviate this analysis by asking a substantive and
symbolic question: In the highly automated society
who will change the light bulbs? That is, associate
each of us with a large number of black electronic
boxes that do virtually everything automatically, eliminating the need for unskilled labor, and ask who
would interchange the boxes when one of them malfunctions or wears out? Perhaps the ready answer is
that this also can be done automatically. Then we must
assume the design and building of still more electronic boxes to maintain the first set and these must
respond to automatic diagnosis devices, calling out
automatic putting in of spares and the automatic transportation of equipment from one place to another.
Obviously, the extreme of trying to do away with man
entirely is as silly at the level of the unskilled as it
is at the skilled level. Nor does it make sense to design
a way to operate our society to provide for all of our
physical, intellectual, and cultural needs without intervention by man.
A city consisting entirely of trained engineers and
scientists might either have to remain dirty, or depend
on a means of drafting creative people for an hour or
so a day to keep it clean. Or else the skilled workers
might have to put their time and resources into designing a city that keeps itself automatically clean,
even though this might be so expensive in the use of
their time and resources as to prevent their realizing
many other more important benefits.
The intelligent means of accomplishing that which
man needs and wants done in society is to use an
optimum partnership of man and machine at all levels
of skill. This most favorable condition may be difficult to reach in the future because the percentage of
unskilled workers who will be matched to the duller,
mundane, less intellectual task will be fewer than the
demand. As society moves forward, as we. broaden the
spectrum of man's activities as a result of his being able
to make use of machine partners on the intellectualinformational front, and as more people become educated, we may find ourselves forced to a greater than
optimum reliance upon the machine just because of
a shortage of unskilled labor.
IS TOO RAPID A RATE OF CHANGE
DISORGANIZING SOCIETY?
It is becoming common for all of us to complain that
we can't do anything these days without a huge exercise

in frustrating arrangements. Life is becoming too complex and we associate this with its being one of rapid
technological acceleration. Because earlier societies
were simple and understood, they appeared well or:ganized. The computer age seems to be headed for increasing confusion.
Fortunately, the computer is the foe of disorganization and chaos. It is the tool of all time for carving
orderly patterns of control. Indeed, the computer has
just arrived in time. Electronic information handling
systems are being developed and installed just fast
enough to prevent our being completely drowned in
a sea of red tape. The problem of keeping track of
everything that has to be included to keep the operation of the world running is growing, and the ability of
electronic systems to help us keep our heads above
water is timely and fortunate. Electronic systems are
ideally suited to gathering information, assembling
facts, applying logic, and controlling the flow of all
needed data and directions. The constraint to smooth
physical operations of our civilization today is a
bottleneck of paper, of information handling. It has
been bad enough in the recent past but, in the future,
without modern computer networks, none ,of us would
be able to get our pay checks, keep our insurance
policies active, obtain our bank statements, deliver
messages, keep track of who owns what, and maintain a semblance of order. Without electronic systems
now being developed and implemented we would have,
in short, a much lower standard of living and approach
the very chaos we fear. The misconception here about
the impact of the computer is a sort of "guilt by association." But the computer is the hero and not the
culprit, the defender and the hope, not the attacker
and the villain, of our fast-paced, increasingly complex
society.
WILL THE FUTURE CONSTITUTE
A ROBOT CIVILIZATION?
It is becoming increasingly common to believe, with
resignation and chagrin, in the certainty that ours will
become a robot civilization. Man is envisioned as becoming, in the future, an anonymous cog in a vast
interconnection of cables, computers, signals, and moving vehicles. The world is pictured as a place where
every action of society to its infinite details will be
planned and controlled, with a man a mere number,
an apathetic, nonparticipating disinterested bystander
in decision making.
Such a structuring of society is inconceivable unless
one simultaneously postulates the existence of a pervasive, automatic electronic information system that

The Computer and our Changing Society
senses all of the data needed to control our daily
operations; one that is busily engaged on a mass scale
in communicating information, to which machines and
men all respond. Now, if such a system were to exist
at some future date, it would make possible a mode of
social operation which would be quite the opposite of
robotism. A ubiquitous electronic information network,
as effective and efficient as this, would make possible
the disseminating to every citizen of all items of information worth evaluation for the selection of goals
by the people. This network could gather opinions and
use this polling in the making of decisions.
The technological advances that one must assume to
be concerned over a future robot society would also
make possible individual participation in our homes of
a form of "instantaneous democracy." The same system that can tell millions of people exactly what to do
can just as well ask them to choose what to do from
a group of well-presented alternatives. The citizens of
the future, so far as technological potentials are concerned would be able to tune in on the highest level
discussion of the big issues and take part by expressing
their opinions electronically from their homes in the
deliberations of the Congress, state legislatures, and city
councils. I am not suggesting that it would be to our
advantage to have every citizen share in every decision
that affects the complex operations of our nation. But
the same technological system that makes possible a
robot society, where everything is controlled because
all the information needed for control is at the right
place at the right time, also can enhance democracy. It
can make possible an informed, interested public, the
tapping of citizen opinion on issues, and the creating of
vast loops of citizen participation in decision making.
Whether we move towards a robot society or in the
direction of "on-line democracy" is not determined by
technology. Science merely offers us the choice. A misunderstanding of the possibilities, a firming up and
prior acceptance of the myth of robotism m·ay keep us
from having a choice.
WILL FREE ENTERPRISE
BECOME EXTINCT?
The advent of the computer age appears to many to
carry with it the dreaded planned economy. They
forecast an automated socialism without free enterprise or private capital at risk, a detailed control of
the economy, meaning "state control." Their fear is
that we will lose creativity, individual initiative, the
advantages of competition and incentives and, especially, give up the freedom to take individual paths to the
new heights which. men can attain when unhampered

and not overly controlled. Nor is it consoling that the
Soviet Union's attempts to completely control her
economy by planning from above is falling short of her
goals in practice. The fear is that perhaps the Soviet
Union tried to do it without the means; with a broadly
based network of electronic information systems it
might become technically practical. Then if the government and the people choose to do so, they can arrange
for this kind of a controlled society in the future.
To many individuals there appears to be only two
paths. One leads to planned economy, the socialistic
state. The other winds back to 19th century entrepreneurship, so important in building America's economy but hardly accessible to us now. But it is submitted that there is a third, much more likely path. It
leads to the creation of a free market of an unprecedented form and level.
To perceive this route we need to note first that
planned economy cannot really control in detail unless
the consumer is in the loop. Perhaps the Soviet Union
has proven this point; although the government can
plan what to produce, it cannot force the citizens'
utilization of the products in accordance with the plan.
As we try to do this in the technological society again
we must assume the existence of an all-embracing
electronic information system that can reach every man
and machine. If we can communicate with everyone to
ensure the working of the economy, as planned, we can
also ask everyone what they want out of available
product possibilities. We go on now to note, oversimplifying perhaps to make a point, that the essence
of true free enterprise and our capitalistic system is the
free market. So long as we have a means for people to
freely choose on what they will spend their money, so
long as the producer is able to offer his ideas and goods
publicly for sale, then we will continue to have the
advantages of free enterprise. It is only a detail how
automated the flow of information is, if that information
flow is used to step up the process of consumer selection,
capital investment, production, and distribution.
So the basic concept is that a national electronics information system that has the technological capacity to
effect a thoroughly controlled eco!}-omy in principle,
run from the top (the government), must also provide
the necessary communications for a vast consumer free
market network run from the bottom (the buying
public). In fact, a nation which has a working electronic
information network reaching every nook and cranny
of its economy can create a free market of a form and
on a level that civilization has never previously known.
It could be a market in which everyone knows quickly
what is available. A proposal to produce something of
interest to possible purchasers could be quickly viewed

The Computer and our Changing Society

and assessed by potential buyers. Each of us in our
homes could electronically respond directly to a "commercial" that describes next year's contemplated automobile models and offers a substantial discount for
orders placed now. We could step to our sets and push
the right buttons to confirm our purchases. This kind
of direct consumer information applied to vacations,
houses, soap, refrigerators, and even educational
courses could be used by the automated network to
schedule in detail, from the ordering of basic raw
materials to the setting up and manning of plants and
facilities. Planning and control is not practical if based
on scheduling from the top. But it can work if based
on commitments by the buying public that can be
followed through, and on the basis of which plans can
be made with confidence. With a potential free market
which could be made to exist by the year 2000 in an
automated, electronic, rapid, all-embracing on-line
form, planning that includes the consumer would be
possible by the entrepeneur. This is an entirely new
form of free enterprise, different and quite superior to
the 19th century form. Capital investment in an environment of a fast-responding customer, an "electronically" informed, interested, active market, makes
possible increases in the profit-to-risk ratio. Given
the opportunity to participate in an accelerating free
enterprise which advanced technological systems make
possible, then we might expect people to enthusiastically
pursue this path. Ideas for new products quickly disseminated would beget other ideas. Efficient scheduling
for production and distribution would leave more resources for risk-taking.
It is equally important to note that a strong role for
the government will exist. This is to provide service
and to referee, control and assure objectivity, honesty,
and opportunity in this free market which depends on
such a huge national electronic information service. The
government will be so busy impleme 'ting this service
and it will be under such pressure f:"Jm the voters to
further expand individual participation in the free market, both as consumer and risk-taking supplier--everyone will want expanded opportunity to participate and
benefit-that we will not need to worry about the
government's seeking to plan the economy from the top.
In the future, technology does make possible a kind
of automated socialism, a regimented, governmentcontrolled economy, poor as that might be. But it also
makes possible a government-aided, unprecedented

level of genuine free enterprise with creativity, incentive, competition, and individual initiative carried to a
new golden era of opportunity for man.
THE IMBALANCE OF TECHNOLOGICAL
AND SOCIAL ADVANCE
The growing myths and misconceptions about the
impact of computers and other advanced technological
additions to our society have their overall foundation
in the mismatch between rapidly accelerating scientific
progress on one hand, and lagging social advance on
the other. This imbalance is shown in our having developed an ability to release tremendous amounts of
energy virtually able to destroy civilization before we
have social maturity sufficient to preclude this possibility. We are on the frontier of radical advances in
biology that can conquer diseases and prolong human
life, while still socially unable to handle birth control,
and thus we must be continually concerned with the
problem of population explosion. Our large space program, justifiable as it might be for research and commerce, arose not because of an appreciation of these
factors by the citizenry, but because of a prestige race
with another nation. Finally, we have the possibility
through automaticity of providing easily for man's material needs, but we are afraid we may not be able to
handle the subsequent dislocations. It is these inconsistencies resulting from our socio-technological imbalance that are producing problems today, and not
scientific progress.
The computer is not the source of imbalance, it is
a tool that can accelerate civilization's progress and
bring technology and society into alignment. Computers give us more brainpower. Properly used they
can help us increase our natural brainpower by improved education. They can increase our utilization of
man's mind. They can give us tremendously greater
informational-intellectual capacities. A man-machine
partnership, with the computers handling the mundane,
rapid processing of data and providing instantaneous
display of information where it is needed, allows man
to rise to the higher intellectual pursuits. That is, the
computer will make us smarter. Perhaps, we will even
become smart enough to broaden our humanistic perspective. Then we can apply scientific innovation universally for the moralistic as well as economic good of
our society.

Computers and education

R. W. GERARD
Dean of the Graduate Division
University of California at Irvine

use in at least the developmental, if not the completely
operational, stage.
Actually, the problem clearly before us today is
minimally that of hardware and maximally that of
software. This is discouraging to many people. It may
take as much as 200 hours of an expert's time, with
some additional programmer time, to program one hour
of effective tested-out computer-aided instruction.
This seems at first, indeed, a devastating problem; but
second thoughts are more hopeful. The languages for
programming are still relatively primitive, although they
are approaching basic English, and will become more
efficient. But even if they did not, please consider that
since the invention of the earlier technologies of communication-of which language itself is the main oneit took several millennia before man achieved writing,
let alone paper and printing; and the software composing a modem language, English or any other, and
the entire literature now available in that language,
could be replicated on computers (reduced to computerized handling, put in appropriate memories, and
so on) in something like 10-2 as great a time-and its
usefulness probably be made 102 times greater. A gain
of 104 is worth the time and trouble and, as I hope to
show, there should be adequate funds to do the job.
Incidentally, as Dr. Ramo pointed out, people fear
that computers will replace them or, as someone put
it a few years ago, that artificial intelligence will replace
natural stupidity! What we are really facing, of course,
a symbiosis of both, combining the attributes of great
speed and vast memory of the idiots that we call computer systems with the imagilJative, creative, idiosyn-

Since I am an enthusiast, rather than an expert, in
the computer field and have been asked to talk about
the future, anything may happen; but it is reassuring to
find experts almost as far out as myself. In fact, it
would be unwise to look only at the immediately available technology; and I have often thought of the early
days of aviation, when Billy Mitchell tried very hard to
convince his superiors that the airplane had a future in
war and, as you know, was essentially cashiered out of
the Army. He did induce a cavalry general to give airplanes a try, but the experiment was a failure. One of
the "crates" of the day flew over a polo field while officers were playing and attempted to hit players with
oranges. No hit was scored, which was pretty conclusive
proof that there was no future for aviation in warfare!
When I became actively interested in this problem
something under three years ago as we were beginning
to plan for U. C. Irvine, the experts in the field whom
I consulted were of limited optimism. They said, "Yes,
the things you want to do will be possible in ten years."
Two years ago, taking another sounding, I was told,
"Yes, tho things you want to do will be here in three
years." Last year, at the IEEE, the speaker before me
in effect said, "Look, we have available, now, large
memories with rapid access, parallel processing, mUltiple
access to the computer, improved input/output terminals, improved (richer and easier) programming languages; and I find no roaring demand for them." I was
happy to get up after him and say, rather emphatically,
"The roaring demand is here." Now, at this meeting,
many of the items which even a year. ago were merely
potentially available are on exhibit and are in actual

Computers and education

cratic, pattern-forming capacities of the human brain
and mind. I have felt for some years now that, of the
evolutionary epochs in the rise of man from the time
his brain became good enough to develop a communal
culture and pass on experience, which is essentially
education, the major advances were: first, the ability to
have symbols at all (which may have been pre-human) ;
second, the organized symbols of language; third, the
organized tested symbols which constitute science; and
fourth, the prosthesis to extend the mind which is the
computer. I would put this present technological evolution as equal to the development of language; considerably more than the development of printing.
Now a few words more on the evolutionary aspects
of learning and the role of education. Learning is, in
effect, modifying one's behavior in the light of experience; and education, formal education, is an effort to
organize the experience to which an individual is exposed so as to develop a maximal change in behavior
and capacity along certain desired lines. This is nothing
new nor unique to the formal level; indeed, the evolution of the nervous system itself is very clearly the
consequence of responding to the species' experiences.
There is good reason to think that the cerebrum developed in response to the steady barrage of the brain
by nerve messages elicited at. improved receptors, particularly the distance receptors. Originally smell, later
vision, and to a lesser extent hearing, are the inputs that
give an animal warning of changes in its environment
before a predator is upon its tender skin or before it
flushes its own prey. This maintained input supplied
the exercise which increased the brain in the first place
and, as current analysis of the anthropological evidence suggests, this process was greatly accentuated
when man began to use tools-which rather explosively
further enlarged the brain.
Now, in the development of the brain of the individual
infant, exactly the same sort of processes are occurring.
It has been demonstrated by a rich array of varied
experiments that an infant deprived of certain experience fails to develop the capacities for handling such
experience. A baby chimpanzee kept in the dark, or
even with milk glass over its eyes so that it sees light
but not patterns, until it is some weeks old (nothing
done to the eye, nothing done to the brain) may never
learn to discriminate patterns in the environment. It
remains functionally blind. Conversely, if one enriches
the experience of a young animal, a rat in this case,
there is actually a hypertrophy, a thickening of the
cortex of the brain; just as exercising muscles increases
their size.
So evolution has depended on the impact of the environment on many generations of individuals in the

species; individual development, on the impact of the
environment on the single individual. It is too large a
story to go into here, but as a biologist, I would maintain that the major theme of all biological evolution
has been the increasing responsiveness of organisms to
the environment, the developing of malleability-not
merely being able to learn in response to the environment but learning how better to learn. The same is
true in the education of the individual child; it also
has to learn to learn and can thereby learn better. I
would, therefore, add to the previous statement, that
the computer will be a great supplement or prosthesis
to the human brain, and that the computer can also be
a great developer of the brain. Here is a tremendous
opportunity for the future.
Formal education probably should be traced back to
the 12th century when the great medieval universities
began, led by the University of Paris. We may forget
that the students there were often subteenagers, so a
university wa~ not qnite what we think of today. The
widespread present formal education, essentially in '
reading, writing, and arithmetic, was not only nonexistent but would have been utterly useless before printing
and the widespread availability of books made it possible to read and therefore made it worthwhile to .learn
to write. It is only within five centuries that our current
widespread literacy became possible and was achieved;
even reckoning, especially multiplication and division,
was practically unknown three centuries ago but was a
special skill of highly trained expert clerks. Perhaps
because education got started relatively early in the
area of behavioral institutions, perhaps because of the
inherent difficulty of the subject matter-the vagueness
of the resources and even of the goals, the uncertainty
of the outcomes, and the extremely varied artistic and
uncontrolled methods of procedure-I submit that this
particularly important area of human activity (perhaps
the most extended and continuous and largest activity of
man; wars are bigger but they come and go, science is
more continuous but less extensive), education, has
lagged very far behind other areas of applied behavioral
science. It is not evaluated, it is not analyzed, we really
don't know whether or not we have been really acchieving anything with all the time and funds and human
labor that have gone into this field. The most important
impact on education of computer technology (and I
use this in its broadest sense) will probably be by
supplying a tool for finding out what we are doing,
for turning anecdotal impressionistic answers into
scientifically testable ones, and so turning what has
been almost purely an art into a respectable scienceand without eliminating the artistic aspects either. Research in education, advances in educational under-

Computers and education
standing, and education as a behavioral science will be,
I think, the· most important outcomes.
Let me give you 'a concrete example. There are two
rather important competing theories as to the way
learning occurs, either as a steady incremental process
or as a step-function increment-differential or integral
advances. The two theories make precise predictions as
to the best reinforcement schedule for learning. According to one, those items on which an error occurred more
recently should be presented oftener; according to the
other, there should be no such emphasis and all items
should be presented equally but at random. It would be
impossible to perform an experiment of this sort without the use of computer technology, computer aided
instruction; with this, Atkinson at Stanford is doing just
such studies with children.
Let us, then, examine in a little more detail the impact of this new information handling technology on
the procedures and the institutionalization of education
and try to foresee some of the social outcomes of these
changes. To give you a clearer picture of what I have
in mind, I shall present our activities and goals at UCIit will be clear which are which. The campus became
interested in the possibility of really interweaving these
two great information handling systems. A university is
primarily a system for storing, retrieving, processing disseminating, and creating information (research,
which creates information, being less present in lower
institutions); and computers do exactly the same, even
creating information in the sense that mathematics,
though a tautology, creates usable information by manipulating existing knowledge and assumptions. Clearly
they are made for each other and our hope was to build
a total system for information handling by combining
the resources of both. We set up, for example, a computer "facility" rather than a computer "center," to
imply an interpenetration rather than a boundary.
Although we opened our doors to students only two
months ago, a reasonably powerful computer resource
has been on hand in trailers for a year with the cooperation of IBM. Dr. Tonge, Director of the University
facility, and Dr. Keams from IBM have led this activity
and the development of CAL There are already in
action 18 on-line time-sharing consoles, and a group
of some 20 professional people involved in the computer activities, penetrating all parts of the university.
On the administrative side, we look to a total systems
use in enrollment, in finance, in faculty and student
records, in plant upkeep, in classroom assignments, etc.;
and several lines are now active. Of course, the system
will be used at all levels of research, up to social simulation. systems. The bookkeeping aspects of the libraryhandling of materials, accessions, charge-outs, etc.-

are being prepared for automation and we look towards
processing the information rather than the documents
that handle the information, to get information on an
on-line rather than batch processing basis. Beyond all
these, and our major thrust, has been a concern with
the possibilities of computer-aided instruction in the
educational process, itself; and it is only of this that
I shall "continue to talk.
Let me give you a vision of what is to come, whether
in years or decades remains to be seen. I would suggest, however, that the reason the experts in the field
almost invariably are rather pessimistic in their guesstimates of the future is that, while each is acutely aware
of the bugs that have troubled and delayed his own developments, he is likely to overlook the countervailing
effect of parallel activities by many people trying to do
similar things. The total impact of this collective thrust
is much greater than the· frustrations and delays in the
individual case, so advance moves faster than the general expectation.
I like to think of the total system as a sort of sandwich, of data bank on one side and users on the other,
interacting via the information processing apparatus.
The data bank includes not only the ordinary internal
and external memory devices of the computer but the
mobilization of any kind of material which can be
recorded-on video tapes, movie tapes, micro forms,
slides, phonograph records, thermoplastics, what you
wish. This great data bank need not be located physically in relation to any particular processor but be
made available to many by networks. Such banks can
be accumulated in a very few places (conceivably in
only one except for the danger of loss or damage) and,
by the communication networks, be made available at
least areawide to all sorts of users in all aspects of
education. On the other side, I see the user sitting at a
terminal, a console or carrel, at a convenient location,
including his own home, and able to communicate in
and out not only by typewriter, by cathode ray screen
plus light-pen, and by voice, but also doing these with
relatively simple buffer arrangements and using an adequate communication language. Beside the individual
user, group interaction should be handled-a seminar
with the instructor and a dozen or more students interacting audio-visually via a communication network and
controlled through the computer processor, much as if
all were in the same room.
The processor itself I like to call the tutor because
here is an agent potentially updated in subject matter;
and learning improved heuristics of the precise educational techniques to be used for this or that kind of
individual with this or that previous training and with
this or that temperament-a hostile youngster who likes

Computers and education

to fight back or a passive one who must be coddled and
brought along. The tutor will have built into its own
memory a detailed knowledge of each student with
which it is working, in terms of that individual's background, personality, and achievement in the particular
field; will be able to give the tutee immediate individual
attention and to do so without threatening and with
infinite patience.
Such an entry of computer technology into the educational process will have far-reaching and crucial consequences. I will not have time to go into all of those
but, fortunately, the problems have already been discussed: the danger of regimentation; the danger of BigBrother in "1984"; the danger of depersonalization or
dehumanization or social anomie-existentialists insisting that people will lose their identity, individualityall these certainly need more consideration than I can
now give them. At least note that the book, itself,
interposed between humans has not dehumanized
them, nor has Othello as played by Olivier on the movie
or TV screen; nor has mechanization of the kitchen
ruined the home.
Now a brief look at costs. If it took, say, a hundred
hours to program one hour of computer-aided instruction and this program were used by a hundred different
teachers only once, the time cost would balance. In
terms of financial cost, I have picked up a few figures
here and there. For example, in the elementary grades
it costs 27 cents per pupil per hour for teaching (this is,
I believe, a national figure; the cost in California is
nearly double); the work (including all development)
cost $1.00 per hour per pupil per simple terminal according to Atkinson. But vastly more is done in that
hour; in fact, it is reported possible to teach a child to
read in 200 hours on a terminal-a cost of something
like $200 per child! Since only a seventh of teacherchildren contact time is spent in teaching, say five
hours a week, it is easily understandable that great
savings in student hours are possible. Further, if one
teacher taught 25 children to read during one full yeaf-,
this would cost well over $200 a child in her salary
alone. There is much reason' to believe that we could
squeeze as much as three years out of the K to 12
period of schooling and not leave out anything of
worth. In effect, during the 10th, 11 th, and 12th years
students are doing nothing productive in society and
are costing a great deal of money; cutting these years
is estimated (Machlup) as giving an annual saving of
$15 billion. The cost of computerizing the whole of
education, bringing all the resources-all libraries and
everything else-into a machine-handable form, building the necessary programs for very rich Socratic tutorial interaction with students even at fairly high levels,

would be paid for in very few years.
At the university level, if all undergraduate teaching were done solely on terminals (which, of course,
would not be for a long time and probably would never
be desirable), and if one achieved a gain factor in time
of three to one (a gain of over five to one has been
claimed in some early "Studies in this field), four courses
at present requiring three hours each a week would
need four hours a week on one terminal per student.
If a terminal were used only 40 hours a week it would
service 10 students; actually, longer use is easily obtained; so, say, all the work of 15 students could be
carried on a single terminal-for 10,000 students less
than 700 terminals would be sufficient for the whole
educational process. At a student-faculty ratio of 15:1
and an average salary of only $10,000, teaching cost
alone would be nearly $7 million for 10,000 students.
The largest present hardware systems rent for $0.5
million a year and require a like sum to staff; double
this for capacity and add another $2 million for terminals, and the cost is still half that of human teachers.
Further, if three man-years are needed to program one
course, and 100 courses would cover the bulk of undergraduate needs, all this software could be produced for
some $3 million. Even if obsolete in one year (three is
more reasonable) this cost is trivial if the system is
widely used.
Well, aside from the financial aspects, the advantages
of such a learning arrangement are only to be enumerated to be recognized. For the student it offers: (1)
better and more comfortabl~ and faster learning-he
can time his learning experience at his convenience, go
at his own pace and catch up missed time; (2) better
teaching at many levels and in many areas; (3) particularly important-personalized tutoring, individual
attention (I remind you, what is so often forgotten, that
Thurstone's original study on the primary mental abilities showed ratios as high as a hundred to one in favor
of child A over B for ability 1 and the reverse ratio of
a hundred to one in favor of child B over A-the same
two children-in ability 2; there are clearly fantastic
differences in human beings and it is high time that we
stopped batch-processing them through the educational
machine! Let them take the initiative for actively learning in their own ways); (4) automatic measurement of
progress, by keeping appropriate records of responses;
when the course is finished, the examination has been
taken, and examination neuroses are bypassed; (5) vastly richer materials, demonstrations, exhibits, travel material, on-site work at archeological excavations or
ocean bottom laboratories, are available for presentation; (6) more sophisticated problems can be included
in instruction even to the level of simple research, lift-

Computers and education
ing the drudgery of sheer repetitive computation.
For the teacher, the system: (1) also takes away a
great deal of drudgery and repetition; (2) allows the
teacher herself or himself to be updated effectively,
without allotting a summer or a year for subject matter
refurbishing every three or five years; (3) encourages
frequent changes in the actual material used; (4) makes
much more time available for real teaching-recall the
estimate for grade school that teaching contact between
teacher and student averages less than 15 % of the time
they are together.
Besides student and teacher gains, there are wider
social goodies: (1) the very best materials can be produced by master teachers; (2) these can be used widely
and repeatedly at least for a limited time; (3) individual
modifications in the program can be made almost at
will, in contrast to the delay and pain of a new edition
of a book; (4) the great information systems can be
tapped freely; and (5) perhaps the most important of
all, the desperate shortage of teachers can be relieved.
In "teachers" I include good teachers or average teachers or almost any warm body (physically, not psychologically); there just cannot be enough humans for the jobs
to be done. Think, further, of the teaching needs of
the emerging countries which must get themselves instant education. If one teacher can teach a few dozen
students who then become teachers who can teach a
few dozen, the multiplying or avalanching effect is just
impossibly slow; so that some of these countries are
making a quantum jump into advanced technologiesjust as some countries earlier went from bullock cart
to airplanes and bypassed the wagon and automobile.
In our own country, autistic children-who rock all
day, insulated from the world-have been led into
participation by interacting with computers when even
master teachers have failed (nonthreatening objects, infinite patience with repetition); and the vast needs of
our Headstart program, to bring mere symbolic thought
to underprivileged babies now growing up practically
without language, cannot be approached with the supply
of human teachers present or future.
Even beyond the immediate teaching possibilities are
the exciting social outcomes for education. I shall
briefly mention three. First, there will be greatly increased flexibility not only in handling the individual
child, but also in handling the materials. If one can
break the teaching sequences into relatively small units
then, as with a Mechano set, a few kinds of pieces can
build a great variety of structures. Instead of having
many different courses in statistics-one for psychologists, one for biologists, one for public health, one for
engineers, blocks of basic elements and particular uses
can be put together in well-tailored fashion. This also

means that it should someday be possible to get away
from the lock step of the semester or the quarter system. Large course blocks should become relatively
meaningless as each student goes through a learning
experience cut to his shape more or less continuously.
A particularly important consequence could be the
separation of the function of certification from the function of education. They are now part of the same process; and getting good grades, in order to enter the next
school and get good grades in order to enter the next
one above, has so come to dominate the whole thinking
of the students and the teachers that whether one gets
an education or not has often been pretty much forgotten. Progress can be certified and mastery tested in
a much more certain and objective way with these
newer technologies, leaving the process of real education (and with teachers used in it especially) to occur
as needed.
Another important outcome, already implied, is the
possibility of spatial dispersion of the learning experience. If certification is covered, then actual class attendance to learn becomes immaterial. I strongly expect
that, in the not-too-far future, there will be an opportunity for the individual to interact through a console
with the great array of knowledge and even with other
humans, so there may not be geographic entities like a
campus. I am well aware of what this means-you
cannot have football and orchestras without coming
together; in fact, it has even been stated that "it takes
two to tango"-but for dealing with ideas, physical
contact hardly seems needed; and high-level seminartype interaction is possible over video conference hookups.
Teachers have been doing a great deal more than
merely help develop the information processing capacities and resources of their students; they have been
friends, they have been hero figures, they have supplied
motivation, they have been shoulders on which to cry.
All these things are very important but perhaps they are
not all functions of the same individual at the same
time; so I see the possibility of again splitting the
separate roles. Indeed, the medicine man once served
his primitive community as priest, lawyer, doctor, teacher, and entertainer; these have been separated into
different professions, but have sort of collapsed back
together in the teacher and now may again be separated
out. In fact, I think it likely that a very different kind
of good teacher will come into being and that new
sorts of people will surge into the field of education.
The great teacher of the future is likely to be more like
the author or the composer or the director than he is,
as at present, like the performer or the actor or the
concert player.

Computers and education

Well, rather than continue about changed content
and goals, let me close with just one thought. Man truly
no longer lives in an outside world. As the cells of the
body in the multicellular organism have created an internal environment in which those cells live, and which
is made possible by the collective action of the billions
of cells, each group performing its own role in the body
organism, so men, in the same way, have created societies, groups of individuals in a multi-individual epiorganism, have created an internal environment of the
society, a culture if you will, in which we live. This is
- also regulated against extreme swings; we are now little
concerned about the physical and biological problems
of the environment, there is no worry about wild beasts,
we go to the supermarket for our food; houses and
clothes protect against weather; we are conquering disease. The pressing environmental problems with which
we live are those that man has created for himself-by
increasing the ease and frequency and range of communication, the number of people who communicate,
and the richness of the material which we now can
communicate to each other. We are rapidly raising a

sea of information in which we must either swim or
drown, and the way we must swim is by enhancing the
problem-solving resources of man and society.
The new computer technology will allow effective
studies of human ecology; of the distribution of, and
environmental influences on, physical and mental health
and disease; of employment needs and their projection,
so that appropriate special training of properly selected
people can produce the round pegs for the round holes.
But of the many opportunities for aiding man to handle
himself collectively, in my judgment, the improved
teaching of the young, to be effective members in society, is the greatest of all.
I cannot close without quoting to you one verse of
my favorite poem, O'Shaugnessy's Ode:
We in the ages lying
In the buried past of the earth,
Built Nineveh with our sighing
And Babel itself in our mirth;
And o'erthrew them with prophesying
To the old of the new world's worth;
For each age is an age that is dying
Or one that is coming to birth.

Computers:
The physical sciences and medicine*
JAMES

MALONEY, JR.

Department of Surgery/Thoracic Surgery, School of
Medicine
University of California at Los Angeles

Although computer science is in its infancy, it has
already contributed significantly to society in the fields
of business and commerce, communication, exploration,
and scientific discovery. In contrast, the contributions
of computers to medicine have thus far been minimal
for reasons which I shall discuss in a moment. It is this
lack of application of computers to medicine which
makes medicine one of the most fruitful areas for the
computer-oriented scientist. I can promise you the rewards will be great for those of you who choose to
apply your talent for the benefit of human health and
welfare. There are two reasons for this:

he be laborer or corporation president, who would
not trade all of his money, business, commerce,
communication, and transportation for 18 more
months of healthy existence. Therefore, the computer scientist who devotes his effort to promoting
human health and welfare will be both generously
supported and greatly appreciated by his fellow
man.
What do we expect computer science to do for medicine? The daily newspapers would lead us to believe that
computers will diagnose disease, store medical records,
interpret X rays, and perhaps remove a patient's appendix untouched by human hand. Perhaps this will
come to pass-but I do not find the possibility intellectually exciting.
Each of you has a much more important attribute to
bring to medicine: the problem-solving technique of
the physical scientist. You would be amused-and I
embarrased-to see a full exposition of the way most
of us in medicine go about the business of scientific
discovery.
The difference in problem-solving techniques of the
physical scientist and the physician is best illustrated by
my favorite analogy. Isaac Newton, after watching the
red apples fall from the tree for some weeks, finally
formulated a generalization which described the behavior of falling bodies. A little boy came up and
said, "That's all right, Dr. Newton, for falling red

1. The systematic methods of scientific thinking
which naturally lead to success in the application
of computers to a scientific discipline have already
been developed, and they have proven phenomenally successful in such fields as high-energy
physics and molecular biology. Discovery in the
field of medicine waits like a ripe apple to be
plucked by the computer-skilled scientist.
2. Society will generously support your efforts. The
value judgment that society places on health and
longevity ultimately is the same as that which the
individual places on them. I have yet to see a
man who is dying of cancer of the lung, whether
*This work is supported by grants-in-aid from the United
States Public Health Service (HE-05357, GM-13242, and
FR3).

Computers: The Physical Sciences and Medicine

apples, but how fast do acorns fall?" Newton- is said to
have given the boy the formula D== lhgt 2 and said,
"Run along and figure it out for yourself, sonny." A
physician, to answer the problem about falling red
apples, would instead have gone to the store, bought a
bushel of apples, a stopwatch, paper and pencil, and
a slide rule. He would have dropped the apples one at
a time from the top of the apple tree and carefully observed and recorded the time it took them to reach the
ground. At the end of that time he would have been
able to say that it was 2.3 seconds with a standard deviation of ± .25 second. If the same little boy had come
along and said, "That's fine, Doctor, but how fast do
acorns fall?" the 20th-century physician would then
have run down to the store, bought a bushel of acorns,
another stopwatch, paper and pencil, and repeated the
same procedure. The modern physician is inclined to
gather masses of random information without ever defining his question and its answer in exact physical
terms.
As Vincent Dole of the Rockefeller Institute so colorfully puts it, "Physicians are obliged to swim in an
ocean of detail, surrounded by great truths in high
dilution. "
The physician already has available more facts than
the human mind is capable of bringing to bear on the
solution of a single patient's problem or in the performance of a single medical experiment. "Like gold in
seawater, facts lose value when mixed with irrelevancies
and can be recovered only by special techniques of
analysis. "
It is the conceptual approach to problem-solving
which I believe will be the computer scientist's greatest
contribution to medicine. John Platt of the University
of Chicago has attributed the stunning successes in the
fields of high-energy physics and molecular biology to
the regular, systematic application of Francis Bacon's
system of inductive inference to the design of experiments. The physician has much to learn from the
physical scientist in this regard.
There are many impediments to the introduction of
computer technology into medicine:
1. There is a communication barrier. Unfortunately,
the biologist is poorly equipped by virtue of his
education and thought patterns to make use of
the physical sciences.
2. Physicians are understandably fearful of machines
and methods they don't understand.
3. There are limitations in both hardware and software. For example, we can train a medical student to recognize appendicitis, but we can not
train a machine to do so.

It is well recognized that appendicitis is frequently
characterized by nausea, abdominal pain, fever, and a
high white blood count. And yet a medical student will
not fail to diagnose appendicitis even if only one of
these four cardinal signs is present. On the other hand,
he may make a correct diagnosis of intestinal flu even
if all four of the signs of appendicitis are present.
Twenty medical students may all arrive at a correct
diagnosis of appendicitis, but each one may do so by a
different logical pathway. In contrast, we seem to be
able to approach the problem by computer only by (1)
Beysian statistics, which give us a probability of the
diagnosis of appendicitis; or (2) sequential branching
logic, which usually leads us to the wrong diagnosis.
-Several problems with our hardware and software
seem to be:

1. Computers do not have 1012 bits of storage like
the human brain.
2. Computers do not have redundancy like the
human mind.
3. We do not understand the pattern-recognition of
the human mind sufficiently to be able to duplicate it in software.
Neverthless, I am encouraged. The educational background of our medical students is improving tremendously. For example, in our current freshman medical
student class, 90% of the students have completed
undergraduate courses in calculus. Unfortunately, these
freshman medical students just out of college have another four years of medical school, another five years of
postgraduate training, and. sev~ral more years after
that before they be~ome productive and 'begin to reap
the benefits of their physical-science orientation.
In our Department of Surgery at UCLA, we are
overcoming these impediments, bringing together both
physicians and physical scientists to work in cooperation on common problems in a medical environment. A
pilot program has been established with Drs. Edward
C. DeLand, James DeHaven, and Norman Shapiro of
The RAND Corporation. By supporting this program,
The RAND Corporation has given us in medicine a
unique insight into the potential contributions of the
physical scientist to biology and medicine. We have
taken engineers with advanced degrees, brought them
into the University hospital, and started them on the
road to a Ph.D. degree in biological science. We are
introducing physicians who have had from two to five
years postdoctoral training in medicine to mathematical
modeling and other computer techniques. The physical
scientists and physicians work side by side in the same
laboratory. In this way we achieve some of the same

Computers: The Physical Sciences and Medicine
benefits as if each individual were educated in both
sciences. Thanks to the progressive attitudes of the National Institutes of Health and its civilian consultants, it
has been possible for us to fund several million dollars'
worth of hardware and other facilities to bring medicine and computer science together in a hospital environment. Although the program is new and it is too
early to judge it critically, we are tremendously ex-

cited by the strides that have been made by bringing
computer science and medicine together.
During a recent 36-month period, 4 Nobel prizes in
physiology and medicine and in chemistry, were given
for computer-based discourses. From this evidence
alone it is apparent to me that computer science is
going to be at the forefront of medical discovery in the
coming decade.

Impact of computers on retailing
C. ROBERT McBRIER
-Vice President, Finance,
Woodward & Lothrop
Washington, D.C.

taining a check digit. Since data often has to be referred
to' outside the computer in an alpha sequence, such
as customer records, we are confronted with the need
to establish a numeric system with gaps to realize
both an alpha. and a numeric sequence. With the input
of numbers in a data recorder by thousands by salespeople in an environment that is not conducive to
accuracy, check digits are required to maintain a
satisfactory level of performance.
For years there has been recognition that a "pointof-sale" device for input of' data' relative to a sale of
merchandise is essential to an advanced system. At
Woodward & Lothrop we have installed more than
800 Data Recording Electric Accounting Machines
(D REAM) throughout all of our 11 stores. This
equipment performs all of the functions of the traditional cash register, but with the introduction of stylized font to print the journal tape, a non-add key
for entry of descriptive data, and a few other changes,
an on-line system has become available to retailers.
The journal tape is read each night and the following
data is available for each sales transaction:

My presentation today will be confined to systems in
retailing that are either "in operation or in the development stage. There is little point in describing retailing
as seen through the eyes of, a customer since a store is
a familiar sight to each of you. Retailing as seen
through the eyes of a computer is quite a different
picture however. For you and me, who are interested
in cybernetics, it is an equally exciting one.
I deliberately used the term "eyes of the computer"
since optical character recognition is essential to implementing the concept for use of computers in retailing, just as magnetic character recognition is basic to
banking.
I do not have the time to round out a full concept
for use of a computer in retailing in the half-hour
allotted to me, so with your indulgence I will refer
only to the major systems.
There currently are very few "real" time applications in retailing and because of the economics involved
there is primary interest in on-line systems supplemented
with some real time. The term "right" time is now being
used to identify this combination of systems.
The primary files in use are:

Salesperson's number
Employee number-for Calculation of discount
Department
Class
Item number
Amount
Customer number
Type of sale

Item merchandise
Resource
Employee
Customer
Financial data
These files are addressed by numeric codes con-

Impact of Computers

RetaUiog

There no longer is any limitation on salespeople or
the merchandise they record in a register.
Department, class, and amount are entered in one
pass. Other data, such as salesperson number, customer, item number, is indicated with a supporting
code number and the data key.
The equipment is not limited to the entry of sales
data since it acts as a communications device from the
sales area to the central information center. Other data
such as the following can be entered with appropriate
codes for identification.
1. Time of day-to determine sales by periods of
day.
2. Stock counts-to update perpetual inventory
records.
3. Employee time of signing in and out-for payrolL
Variations of this equipment include the capturing
of data on magnetic tape, punched tape, or punched
cards in a central area and punched tape at the pointof-sale.
With the installation of these registers supported by
computers, Woodward & Lothrop overcame the hardware limitations for applying management sciences to
retailing. Effective supervision and discipline now remain the limiting factors. In order to implement these
we make available a report daily identifying by registers
all salesperson errors. The computer currently performs 36 audit checks and as additional data is introduced the audit routines will be increased.
I took the time to describe the "point-of-sale" device in some detail since it does constitute a major
breakthrough for application of computers in retailing.
I would now like to direct your thinking to the files
that will be maintained.
ITEM MERCHANDISE
No one system is adequate for the management of
a department store inventory and we often characterize the item as being one of the following:
Staple
Fashion
Big Ticket
Mixed
Generally speaking, systems supporting different
classes of items will include the following:
Staple

1. .Mathematical models are developed from past
history of the same or related items which can

be used for forecasting sales to support an order.
2. Sales activity developed from entry in the "pointof-sale" device, periodic stock counts or print
punch marking tickets.
3. Purchase orders prepared by computer on a
regular weekly or biweekly basis for delivery
direct to stores or warehouse. In the foreseeable
future there should be machine-to":machine communication. Retailer-to-Resource.
4. A punched card activated marking machine is
projected for the near future. The marking, as
well as a turn around document to support receipt of the merchandise, is then a by product of
preparing the purchase order.
Fashion

1. Criteria is developed by price line, class, and
department from past history to identify if a style
is following a normal pattern and will sell out on
schedule, if it is "fast-selling" arid should be considered for reorder, or if it is "slow-selling" ~nd
should be returned to resource or marked down.
2. File updated daily from sales recorded by item
number in "DREAM" point-of-sale device, print
punch ticket or manufacturer's ticket. The item
number will constitute the computer address,
but all reports will include the full description of
the item understandable to the buyer, vendor,
style, etc.
3. A daily report of exceptions, fast and slow sellers,
is prepared for buyer action.
4. A plan of model stocks by category of merchandise by store will be maintained by the computer
to guide buyer in distribution of incoming merchandise.
Big Ticket

1. A reservation file is maintained by all items by
location, which is interrogated from the selling
floor. When the sale is completed, the necessary
forms are prepared in the warehouse to pull
stock and effect delivery.
2. The report is intended to keep buyer informed of
stock status and sales activity to stimulate action on items that are not moving according to
plan.
The item merchandise files will include a reference
to the resource files, but the latter will be maintained
separately for accounting purposes to effect payment
of invoices, as well as for evaluation of overall profit
performance.

Impact of Computers on Retailing
Resource File

1. Ideally for the retailer, th.e item of merchandise
should be marked by the resource in such a way
that the record of its sale can be perpetuated
mechanically.
Efforts to accomplish this are currently manifest in
print punch tickets that have a uniform format for
manufacturers of category of merchandise, as well as
distinctive resource numbers. Interestingly, the more
sophisticated systems of retailers conflict with this concept, since an item number is required to identify the
sale; and this cannot be accomplished with a group
of manufacturers.
2. In order to simplify payment procedures, retailers and manufacturers have cooperated with
Dun & Bradstreet to develop a manual of distinctive numbers for resources. These are included on their invoices and used by retailers to
address the file.
3. Where the order is on record in the computer,
a turn-around document will explode the programs to adjust perpetual inventory records and
effect payment of invoices.
EMPLOYEE
We are working to reduce the number of employees
in the application of computers but they will always
remain the principal element of success in retailing,
since there is a personal contact at the point of sale.
One file will satisfy all the requirements of personnel,
payroll and the supervisor for evaluation of performance.
Where the individual's performance affects the accuracy of the central information center, error records
will be maintained which will be important in reviews.
The successful implementation of computer systems
in a department store with decentralized input demands
that the highest practical standards of accuracy be
achieved.
CUSTOMER
With the initial application of computers to prepare
customer bills, both retailers and customers were upset
that the customer becomes a number on the records
of the company. I well remember the letters we received at the time of our conversion. Some even
punched holes in the tab card or tore it up to show
their frustration.
Our plans for the immediate future are exciting
because we are talking a personalization of the file

beyond what we could ever achieve with clericals.
How does this sound?
1. At the time a customer opens an account she
completes a questionnaire from which we build
a profile of the family. If the questionnaire is
not completed, then the computer will build a
similar record from purchases.
2. The detail of the individual's purchases is identified by store, department, class, item number,
price, and possibly size. The customer number
is also input in DREAM.
3. The detail is stored for billing, but also summarized to indicate categories of merchandise
and price lines purchased. The proper coding on
the descriptive bill will enable the store to be
selective in inserting direct mail advertisements
that should have special appeal. Feedback of
sales should maximize the effectiveness of promotions.
4. A profile of payments on account enables us to
personalize collections and make decisions relative to authorization of purchases. We propose
within two years to interrogate the computer
from the selling floor by telephone and authorize
purchases with a voice answer.
5. As complaints on service are received they can
be accounted for and controlled to enhance
service.
6. As we desire to follow-up on anniversaries,
birthdays, big purchases, specific items, this can
be accomplished.
FINANCIAL INPUT
Planning, control, and evaluation of performance
are, of course, basic to the success of a business. We
accept the fact that from the records of past performance, dynamic plans can be made; from the daily record
of activity, effective controls can be maintained and
management will be able to evaluate performance from
reports prepared' on an exception or other specific basis
to inform the executive.
The timeliness and selectiveness of reporting has been
immensely improved with use of the computer, but we
look for exciting innovations in this field. As the
effectiveness of management can be enhanced, this may
well be the most significant contribution of the computer
to retailing. We now accept the man-machine relationship.
Our immediate plans include variable budgeting. Expenses are now being reported compared to plan and
last year. We propose to identify the extent to which
each account is fixed and variable, and the cost re-

Impact of Computers on Retailing

lated to workload beyond the fixed point. When this
is accomplished, we propose to have the computer
develop budgets based on proposed volume and then to
determine the percent of budget realization based on
actual performance.

Credit

DEMISE OF INEFFICIENT RETAILER

Simulation

The computer will not accomplish for all retailers the
benefits noted. For many years I believed the computer
might help the independent retailer to be more competitive. I now appreciate that many of the concepts
referred to are not available to the small retailer because
he does not have the finances or the staff to implement
them. Service bureaus and arrangements to share computers with banks or other businesses will help, but
they can never match the efforts of the big chains or
the large independents. The computer may well then
speed up the demise of the small retailer, except as the
advantages of the computer are offset by superior personal management. Certainly the future of the complacent, inefficient store is questionable.
Retailing however will get a bigger share of the customer's spendable income.
PRIMARY IMPACT OF COMPUTER
ON RETAILING
I hope you share my enthusiasm for the use of
computers in retailing, even if my reference to specific
applications has had to be brief. In the few minutes left
to me I would like to summarize this impact.
Inventory A1anagement
1. We will operate with lower stocks and achieve
a faster turnover.
2. Balanced assortments will be maintained with
fewer stockouts.
3. Markdowns will be reduced and sales increased
as a result of the above.
4. Buyers will be relieved of clerical functions to
direct their atteniton to creative activities.
Customer
1. As profiles are maintained for each family, we
will be able to render a more personalized service which is vital when the same merchandise
is offered by several retailers.
2. Promotions will be more personalized.
3. Retailing should realize a bigger share of
spendable income.

1. City-wide numbering is now developing which
will be a future basis for combined authorization,
billing, and collection, with material savings in
postage and duplicated effort.

1. As decision rules are defined and systems developed, we will simulate performance to determine the level of customer service we desire to
achieve related to:
a. Inventory investment
b. Achieving ideal schedules for salespeople
2. Through use of management games, executive
training can be expedited and an earlier indication of executive ability determined.
A1anagement, Planning, Control and Evaluation
1. Centralization of functions with improved effi-

ciency at lower costs will be realized.
2. The effectiveness of each executive will be materially enhanced.
3. Share of the market and net profit will be improved.
4. With higher earnings on invested capital, retailing will be more attractive to the investing public.
THE CHALLENGE
The promise of ED P in retailing exceeds almost
every other business and the problems of implementation are just as difficult. Ours is a business of peoplesal~speople, buyers, clericals, executives, warehousemen, delivery drivers-trained to buy and sell merchandise and services to other people-the customers. All
of these people resist ,regimentation and covet the independence to adjust to current situations. Our success in the past has been due in large part to the
flexibility and judgment of these people.
The application of the computer demands that certain procedures be executed with a high degree of
accaracy. The term discipline is now being referred
to more and more as necessary to achieve higher
standards of performance. As the management of a
store understands the potential for computers and gives
leadership to the changes required, it becomes an exciting experience to participate in a team effort where
the results are impressive. With other immediate problems demanding attention, management effort is often
directed elsewhere and the progress is slow.
The large chain has the advantage here since a staff
of experts can be directed to develop pilot operations

Impact of Computers on Retailing
in stores, which can then be extended to others. Enlightened management in a large independent store however can achieve ever more impressive results.
A premium is developing for the quality of leader-

ship, knowledge and ability represented by you in attendance here today to accomplish the improved performance in retailing through use of computers that
can be realized in no other way.

The application of computers to
domestic and international trade
WILLIAM I. MERKIN
and
RAYMOND J. LONG

United States Department 0/ Commerce
Washington, D.C.
national policy cannot be overemphasized. Timely evaluation of these facts is essential.
It is logical, therefore, that the Department of Commerce should look to automated techniques to satisfy
the need for speed, accuracy, and economy in the acquisition and dissemination of commercial information.
As a result the Automatic Data Processing Division of
the Office of Administration, which is responsible for
planning and implementing electronic digital computer
and mechanical tabulating systems for the DIB organizations, undertook a DIB-wide assessment of potential
automatic data processing applications. This study,
which was completed in March 1965, identified nine
areas as feasible for early automation. These areas are:

INTRODUCTION
The Act establishing the Department of Commerce
provides, in part, that the Department shall "foster,
promote, and develop the foreign and domestic commerce . . . of the United States." These broadly described functions are carried out under the policy
guidance of Assistant Secretary of Commerce for
Domestic and International Business Alexander B.
Trowbridge.
The Assistant Secretary is supported by a Domestic
and International Business (DIB) organizational structure comprised of the Bureau of International Commerce, the Business and Defense Services Administration, the Office of Field Services, the Office of Foreign
Commercial Services, the Office of Publications and
Information, and the Office of Administration.
In the performance of their assigned functions, these
D IB organization units must generate a wide variety of
economic and commercial data in order to meet the
increasing needs of government and business for timely
commercial information. Trade regulations, commercial
trade (both foreign and domestic), balance of payments,
world markets, and the problems of businessmen and
investors are of concern to the Department. The importance of gaining access to pertinent facts affecting

American Traders Identification
Foreign Traders Identification
Dispatch Loan Service
Unit Value Index
Export License Application Control
Tailor-made Foreign Trade Statistics
Economic Research
Automation of Domestic and International Business
Publications
Statistical Data Bank
These systems do not employ what would be considered new and unique technology. Rather they apply
proven techniques to new areas of data processing.
Four of the systems are now fully operational: the
Unit Value Index, the Export License Application
Control, the Tailor-made Foreign Trade Statistics, and
the Dispatch Loan Service.
Three systems are partially operational: the American Traders Identification, the Foreign Traders Identification and the Economic Research System.
27

The Application of Computers to Domestic and International Trade

The remaining two systems, consisting of the Commerce Business - Daily and the Statistical Data Bank,
are in the systems design stage.
Together these systems represent the nucleus of
what will eventually be an "Integrated Commercial
Intelligence System."
AMERICAN TRADERS
IDENTIFICATION SYSTEM
The Domestic and International Business area has
established a master magnetic tape file of detailed
financial and commercial data on more than 430,000
U.S. business establishments. Recorded thereon is referenceable information about a company's net worth,
number of employees, sales volume, type business,
products, and other business data. This system speeds
up and broadens distribution, upon request, of commercial information, including international sales and
other business opportunities, tailored specifically to the
needs of the inquirer.
There are three basic classification of files for applying this system; the data on a11430,000 U.S. business
establishments (called the master file), the data on
U.S. business establishments interested or engaged in
foreign commerce (called the American International
Traders Index File), and the data on U.S. business
establishments engaged in foreign commerce who would
be willing to carry, "piggy-back" fashion, another company's products into the foreign market, utilizing their
own distribution system and organization for this purpose.
Prior to the initiation of the master file, firms were
invited to participate in foreign trade exhibits based
upon manual search of a limited reference file consisting of information on approximately 5,000 U.S.
business establishments. This was supplemented by
personal knowledge of Commerce employees on firms
which regularly participated in these activities. Now,
by using a generalized computer search strategy based
upon the theme of the fair, a search is made of the
master file which has stored information on more than
430,000 business establishments. Experience shows that
a usual search of the automated file yields on the averageof 4,000 firms per fair. A mailout on one such output showed that 13 % of the selected firms responded
with an affirmative desire to participate. This was slightly more than 500 firms as compared with about 100
firms under the old manual method. As to speed, 99
fairs can be processed in one pass of the file and a
listing of prospective interested participant firms retrieved-listed separately for each fair, based upon
the theme. The pass takes five hours. Previously, it

took from one to six months to prepare a list for each
fair (depending on the theme and the information
available).
The interest of American businessmen in foreign
trade opportunities prompted the U.S. Department of
Commerce, using a computer routine, to inquire of
340,000 U.S. business establishments whether they
currently were engaged in foreign commerce, and if
not, would they be interested in entering foreign commerce. It was estimated that about 30,000 business
establishments would respond. Actually, almost 60,,000
favorable responses were received. Using these responses as a basis, the American International Traders
Index File was established. A second mailout was sent
to those firms which requested volunteer commercial
information. A total response of 25,000 is expected
from this mailout. The ultimate file will carry information on products (by Standard Industrial Classification),
country outlets, foreign trade representation, and other
pertinent commercial information. The information
furnished by responding U.S. companies will be maintained in confidence within the government and will
not be published. This master file will be used for
spotting trade opportunities. In addition, there are
many firms manufacturing products which can and
should be introduced into foreign markets but who do
not have the capital or facilities for sales and distribution support to compete abroad. Certain business establishments already engaged in international commerce
are willing to use their established sales and distribution
organization to carry, piggy-back, expanded product
lines for foreign sales. The challenge is one of introducing these two types. Inquiries from both types of
firms interested in the piggy-back technique can be
matched and satisfied by searching for products and
carriers in this master file.
FOREIGN TRADERS IDENTIFICATION
Closely related to the American Traders Identification File is the Foreign Traders Identification File.
The Foreign Traders Identification File will be used to
develop foreign trade lists; to identify manufacturers,
agents and distributors; to give leads on investment
opportunities abroad; and to assist in our export expansion program by providing rapid retrieval of information on export markets by country, company, and
product.
Using the country of Japan as a prototype, a magnetic
tape file was established containing current commercial
intelligence information on more than 4,000 Japanese
firms. This information covers the name, address, products, size, capital, sales and other pertinent facts about

The App6cation of Computers to Domestic and International Trade
each company. From this an alphabetic listing was made
on all firms; later, an alphabetic listing by product was
produced; and still later, a profile on each firm was
developed; and finally, a generalized search strategy
for the retrieval of information from the file was programmed. Based upon our experience with the Japan
file, systems modifications were made and plans for
phasing in the remaining countries are presently being
implemented. Arrangements are being made for the
collection, coding, keypunching, and verifying of the
input data to the system. Once fully established, the
file will contain commercial intelligence information on
more than 300,000 foreign business establishments.
The search strategy will be tied into the strategy used
in the American Traders Identification File. This is in
accord with our plan for a single search strategy for
all commercial intelligence files.
DISPATCH LOAN SERVICE
The Dispatch Loan Service is a prototype of what
will eventually be a large-scale information storage,
retrieval, and dissemination system. U.S. Foreign service posts prepare dispatches on recent developments in
products and with foreign business firms. These dispatches are sent to the U.S. Department of Commerce
where they are indexed and a country file maintained.
A monthly index of accessions is published and business
organizations may request a loan of the dispatches,
usually through a Commerce Field Office. The entire
operation is currently performed by manual methods.
The accession rate is about 600 dispatches per week.
The automation of this file will be in three phases:
first, the indexing and coding of source documents to
put them into machinable form; second, developing
program routines for immediate use of data; and, third,
developing and analyzing various techniques for application of data in a larger system. The larger system
we are ultimately concerned with is the storage, retrieval, and dissemination of information on incoming
dispatch communications received by the Department
of Commerce. Approximately 2V2 million copies of
communications were received in 1964.
There are some ready-made descriptors to facilitate
coordinate indexing for putting data into machinable
form. In addition to the usual identifying information
such as reporting agency, country, date, etc., we have
the Standard Industrial Classification (SIC) for products, and have, among other possible descriptors, the
"Harvard Marketing Mix" for the Marketing Factors.
The incoming dispatches will be indexed and stored on
magnetic tape. The monthly index of accessions will be
printed on the computer and duplicated by reproduc-

tion techniques.
Dispatches will be reproduced for distribution on a
request-loan basis.
After the file is operative for immediate use, other
indexing techniques will be tried such as KWIC (key
word in context), and consideration will be given to
more sophisticated search strategies, such as use of
weighing factors, links, and roles.
Using the firms that now use the loan service as a
nucleus, a Selective Dissemination of Information System will be established. This information-sharing concept will assure timely dissemination of information to
the business community.
UNIT VALUE INDEX
The Unit Value Index System produces commodity
value indexes for both imports and exports. These
indexes are used extensively by private industry and
government agencies for economic analysis and forecasting, for the study of price movements of merchandise exports and imports, and for analysis of U.S. foreign
trade volumes.
The system is comprised of three phases: the base
weight computation for commodities which is produced
at the beginning of each calendar year, the data tape
which uses U.S. Census monthly commodity value
statistics for the creation of the commodity value data
tape for both imports and exports, and finally the
computation of the indexes for each commodity by
applying the data tape to the base weight tape.
It is important for businessmen to know the relative
price of a particular commodity in the world market.
This cannot be done effectively unless base weights are
used to determine indexes for specific commodities. The
Unit Value Index System does this by applying a computer routine and logic to integrate and synthesize available information.
The information used is published in many periodicals including the Overseas Business Reports. In addition, among the many private and governmental organizations which use the data are the United Nations and
the National Bureau of Economic Research. Proper
and effective analysis of these data assists the U.S.
Department of Commerce's. mission of developing a
more favorable balance of trade.
EXPORT LICENSE APPLICATION CONTROL
The Office of Export Control of the Bureau of International Commerce, administers the authority provided in the Export Control Act of 1949 as amended,
to control the flow of United States exports. Generally,
this control is exercised by the issuance of licenses.

The Application of Computers to Domestic and International Trade

Approximately 700 license applications are received
daily by the Office of Export Control.
The initial phase of the automated system for this
application has been implemented. This initial phase
consists of a complete reporting system (covering License Applications, Rejections, Issuances, Listing of
Company, Commodity, Country, Quantity, etc.) with
daily, weekly, monthly, quarterly and annual reporting
cycles. Since the quarterly reporting cycle is of significant importance, we are conducting a three-month
parallel operation. The parallel run will be completed
on January 1, 1966.
After completion of the parallel operation, the next
phase will be started. This consists of the processing of
licenses by utilizing computer decision tables and other
types of table look-ups.
The total system ultimately will include processing
applications, generating reports and statistics, preparing
pre-licensing lists, printing approved licenses, file maintenance, updating, and information storage, retrieval,
and dissemination.
TAILOR-MADE FOREIGN
TRADE STATISTICS
This system can best be explained by first considering
the manual methods which were used.
The Business and Defense Services Administration
(BDSA) in support of the U.S. Department of Commerce's export expansion program, among other things,
reviews the impact of imports and exports in domestic
business. Statistical indicators of progress and status
in product commerce and marketing provide an invaluable tool for this review. Under the manual method
the average division required as many as 536 different
source documents monthly for study to develop its
analytical product statistics.
Under the automated system, source data are procured from the U.S. Bureau of the Census in the form
of magnetic tapes carrying monthly import and export
information. These tapes are processed through a
tailored routine which produces monthly for each division commodity and product information formats ready
for review and analysis. Quantity and value of imports
and exports by countries of origin and destination is
included in the tables provided, in addition to cumulative data for the year in progress. Volume data for
each import and export classification are also summarized;
This gives the commodity analyst one" source document from which he can gather statistics. Timeliness is
an important factor in all reporting systems. The statistical data are now provided the analyst within one

week after the data are produced by the Bureau of
Census. Under the manual system, it was weeks and in
many instances, months, before the analyst gathered
the statistics for use.
ECONOMIC RESEARCH
Currently, this system consists of a series of FORTRAN programs for statistical computations and correlations. These program routines are primarily used
in the production of an annual economic study of major
manufacturing industries featuring statistical projections
and forecasts on expected levels of activity.
In addition to the FORTRAN programs, there are
several programs for financial analysis. The financial
programs include: companies listed by various search
criteria, report on" past financial data and ratios for
specified companies, it report on the comparison of a
particular company as it performs within its industry, a
report comparing a particular company's sales and
earnings against the industry and the economy, and a
report of up to five companies showing financial and
market information.
Consideration also is being given to the construction
of econometric models to determine what would be
the relative impact on major industries given varied
economic conditions, to analyze cyclical fluctuations in
commerce and marketing, and to provide productivity
projections.
AUTOMATION OF DOMESTIC AND
INTERNATIONAL BUSINESS
(DIB) PUBLICATIONS
The Commerce Business Daily was selected as a
prototype for developing this application because representative problems associated with automating DIBtype publications are featured in its production.
Federal procurement officers transmit information by
wire and mail on U.S. Government invitations-to-bid
and contract awards for publication in the Commerce
Business Daily. Foreign trade opportunities are provided
daily by the Business and Defense Services Administration. The Daily usually has eight pages of trade in:.
formation, of which four pages are a procurement
synopsis. Presently, the circulation averages 18,000
copies.
The purpose of this application is to convert the
publication of the Commerce Business Daily to a computer-controlled printing process. Once accomplished,
this will facilitate the production and maHout of the
publication on a more timely basis, thereby providing
opportunities for more firms to bid on government
procurement opportunities. This application will pro-

The Application of Computers to Domestic and International Trade

vide the improved facilities required by the Department to fulfill, more adequately, its responsibility for
the prompt and comprehensive daily publication of
federal government procurement actions as required
under Public Law 87-305.
This routine input received by teletype will go directly into the computer and be stored; other input will
be edited and keypunched for storage. The computer
will perform final edit, line justification, and hyphenation. The completed and formatted information will
be produced on paper tape which will drive the "hot
metal machines" at the Government Printing Office to
create plates. The final makeup and camera operations,
as well as the printing, folding, and mailing of the Daily
would be performed in the usual manner.
When the Government Printing Office's high-speed
photo-composition production system becomes fully
operational, additional page makeup composition operations could be performed by the DIB computer.
Fully prepared information could be transmitted by
magnetic tape to the Government Printing Office highspeed composer. Film pages for the Daily could then
be constructed as page makeup negatives ready for
conventional composition, platemaking, printing, and
binding. Ultimately, the system could expand to a total
automated concept for all phases, including mailing.
As we have already stated, this application is a pilot
project for several other Department pUblications (not
yet selected) for publication by automated processes.
STATISTICAL DATA BANK
This application consists of developing a data bank
for statistical analysis and data retrieval.
At the present time, the data bank consists of magnetic tape files of stored data covering commercial

information on 430,000 U.S. business establishments,
information on the financial history of 1,000 U.S. business establishments, information on tariff and trade
agreements, and statistical information on U.S. imports
and exports.
Currently, the automated application of these files
fall into four categories; data retrieval, statistical analysis and correlation, statistical research, and file
analysis.
Specific programs include file-profiles, information
retrieval, and statistical applications.
Many of these techniques have been discussed under
other applications so we will not repeat them here.
The Statistical Data Bank Application is the obtaining, developing, updating and utilization of commercial-economic data of a statistical nature or with
the possibility of generating statistics from the data.
SUMMARY
The above described systems represent a cross section by technology and application.
As to technology, the applications include, among
others: information storage, retrieval, and dissemination; photo-composition; data reduction; data processing; statistical manipulation; report generation; and
decision making.
By application, we have a cross. section of DIB
organizations and a cross section of their tasks. The
data are classed into two broad categories: narrative
and statistical. The nine systems described form the
basic ADP pattern. New systems will be tied into
existing systems, and the existing and expanded systems will be extrapolated periodically. Eventually, the
total system will become a mechanized integrated
economic information system.

The role of computers in space explot:ation
C. R.GATES
and
W. H. PICKERING

let Propulsion Laboratory
California Institute of Technology
Pasadena, California
craft, with respect to the economy and· precision of its
design; the performance of the navigation system, in
the accuracy with which the spacecraft can be guided
to the moon or Mars; and the performance of the data
gathering system with respect to the amount of data
which can be gathered and the speed and. accuracy
with which it can be analyzed.
In my discussion I will describe some of the more
prominent ways in which computers have been used
in the space program.

INTRODUCTION
The modern digital computer has been fundamental
to the space exploration program. Computers have profoundly affected almost every aspect of space technology, including spacecraft design, celestial mechanics,
mission control, and the gathering and processing of
data generated by the spacecraft. Indeed, the evolution
and growth of computer technology is suggestively
parallel to the growth in space technology.
Perhaps the most revealing way to examine our indebtedness to computers is to recall conditions before
computers were available. Fifteen years ago our main
project at JPL was the Corporal, a ballistic missile with
a range of some tens of miles. The calculation of its
trajectory took three weeks of intense labor; now we
can compute the trajectory of a spacecraft to Mars in
a few minutes. Formerly, solving a problem in control
stability or structural design involved many hours of
hand calculation and was likely to require extensive
laboratory experimentation; now, many problems in
these fields are solved more cheaply, more accurately,
and more quickly by means of the computer.
In the last analysis, the effect of modem, high-speed
computers on the space exploration program has been
to increase performance-the performance of the space-

TRAJECTORIES AND
CELESTIAL MECHANICS
The most dramatic usage of computers in the space
program is probably in the field of celestial mechanics.
Celestial mechanics is the oldest of the sciences, and
since a spacecraft in flight behaves like a miniature
planet, our trajectories and navigation are based on this
ancient and ancestral science.
Mathematically, celestial mechanics is· concerned
with the solution of Newton's equations. From Newton's time until very recently, solutions were obtained
only with prodigious labor and by means of amazing
and ingenious mathematical contortions. The digital
computer and the impetus of the space program have
profoundly affected celestial mechanics. Computer calculation of the trajectory of a spacecraft such as Mariner IV is quite straightforward, even though the
numerical techniques used are sophisticated. In the
selection of trajectories for Mariner, many hundreds
of trajectories were calculated and analyzed in order
to determine those trajectories which give the best accuracy, payload weight, and picture quality at Mars.
The computer also greatly increases the accuracy

The Role of Computers in Space Exploration

with which a spacecraft such as Mariner can be
guided. For the guidance of Mariner, the orbit is determined from radar data, and a small maneuver to
be executed by the spacecraft is calculated. The computer permits these calculations to be carried out swiftly
and accurately while the spacecraft is in flight; without the computer, approximations of much less accuracy would have to be used.
SPACE FLIGHT OPERATIONS:
SPACECRAFT COMMAND AND CONTROL
The largest use of computers at the Jet Propulsion
Laboratory is in the real-time processing of spacecraft
data for command and control purposes; we refer to
this process as Space Flight Operations.
Our Space Flight Operations Facility, or SFOF, will
shortly contain three "strings" of computers, each string
consisting of three computers, an IBM 7040, 7044,
and 7094. Into the SFOF come spacecraft data obtained from tracking stations located throughout the
world. The data may be from the spacecraft, concerning
either the condition of the spacecraft or conditions in
space, or the data may be navigational, referring to the
position or velocity of the spacecraft. These data are
sent in digital form over telephone circuits to the
SFOF, where they are routed into the computers.
After processing, the reduced data are presented to
scientists and engineers for analysis and interpretation.
The engineers conducting these analyses have access in
parallel to the computer and are able to request in real
time various programs, options, and methods of display.
Real-time command and control centers are also
used extensively in the manned space flight program.
The control center for Mercury was at Cape Kennedy,
supported by facilities at Goddard Space Flight Center. The control center for Gemini and Apollo is at the
Manned Spacecraft Center in Houston.
Two basic types of computer processing are carried
out in Space Flight Operations. In the first type, the
real-time data stream is simply separated, translated
into convenient units and symbols, and presented directly to the engineer or scientist. This is the "real-time
mode," and it has been the backbone of spacecraft
analysis. In the SFOF at JPL, one computer per string
is devoted to this function, which includes sorting, decommutating, formatting, and preparing printer and
plotter output.
In the second type of processing, a quantity of accumulated data are analyze~ and certain parameters extracted. An example of this type of processing is
orbit determination, in which several hours or days

of tracking data are analyzed in order to determine
the orbital parameters of the spacecraft.
We have been one of the pioneers in time-shared multiple-channel computer usage, in which a number of
users can simultaneously communicate with the machine and receive output. At times I think we would
have gladly relinquished the pioneering honor to someone else; the problems in such a system-the interaction
between programs, the difficulty in diagnosis, the intelligibility of a program only to the programmer who
wrote it, the difficulty in reproducing a condition in
which a failure occurred, the problems of meeting
schedules and keeping adequate documentation-are
well known. However, the system has carried out its
flight mission reliably and successfully, the experience
gained is valuable, and most of these difficulties appear
to be behind us.
SPACECRAFT DESIGN AND
SPACE TECHNOLOGY
In a spacecraft such as Ranger, Mariner, or Surveyor,
subsystems derived from a variety of technical disciplines must all function together with tightly knit
precision and harmony. Guidance and control, communications, science, structural, thermal, and propulsion subsystems must work together. Furthermore, in
spacecraft design it is necessary to consider almost all
of the physical properties of these subsystems-electrical properties, weight, mass distribution, heat generation and conduction, reflectivity, etc. Also, there are
overall constraints of weight and volume placed on
the spacecraft. And for a planetary spacecraft there
is a totally inflexible constraint of time; the spacecraft
may be launched only at infrequent intervals, necessitating accurate and reliable scheduling procedures.
The use of computers has increased the accuracy
and speed with which the spacecraft and its subsystems
can be designed. Using computer techniques to explore
a whole spectrum of design possibilities, design parameters can be selected quickly and precisely. The result is a materially increased performance. Without
sacrifices in cost, reliability, or schedule, the applicati~n of computers results in a spacecraft which will
send back more and higher-quality scientific data regarding conditions in space.
As an incidental aside, it is interesting to note that
our Mariner-Mars spacecraft contained four digital
computers. These were used for science data handling,
engineering data handling, command processing, and
command sequencing.
The impact of computers on engineering technology,
in fields such as structural design, circuit design, heat

The Role of Computers in Space Exploration
transfer, etc., has generally followed a pattern of great
interest and importance. Initially, computers were used
to solve problems in their traditional form. For example, the solution to a differential equation would
be obtained by hand in series form, and the series was
then evaluated on the computer. In the next stage of
development, the computer was given the differential
equation directly, which was less work for the engineer
and usually yielded a more accurate solution. Today,
it is common practice to state the problem to the computer in terms of the end result desired, and the computer both formulates and solves the differential equation. Thus, the computer has been applied further and
further upstream.
CONCLUDING REMARKS
In summary, we find that the computer has touched
almost every phase of the Space Exploration Program.
Spacecraft design is now based on computer analyses,
and the result is a better spacecraft; furthermore, a
steadily increasing portion of the design process is
being carried out by the computer. In celestial mechanics, the effect of computers ..,has been dramatic.
And in space flight operations, all data gathered by the
spacecraft are processed by computers Indeed, the number of computers we will shortly have serially in the
data stream is startling; counting computers in the

spacecraft, at the tracking sites, in the communication
link from site to SFOF, and at the SFOF, we will have
up to 10 computers handling the data.
The role of computers in semitechnical and nontechnical areas has also been vital. Computers are
used for configuration management, budget preparation, inventory control-the list is endless.
Turning to the future, we are looking forward to the
attractive features offered by the new generation of
computers-the much-needed increases in capacity,
speed, memory, reliability, and ease of programming.
However, the impact of computers on basic technology
is likely to have the most significant future effect on the
space program. When I was a student, a large part of
the technology was devoted to problem-solving methodology. Today much of this methodology is becoming
obsolete. The new methodology, based on fundamental
principles and the computer, makes possible the ready
treatment of larger systems. Instead of having to analyze
the relationship between resistor A and resistor B, the
engineer may treat the relationship between circuit A
and circuit B, or system A and system B. The viewpoint is less microscopic, and more macroscopic. For
example, we are beginning to treat system characteristics, such as cost and reliability, as design parameters.
Thus the symbols by which the technology is represented, and hence our way of thinking, are being
changed by the modern digital computer.

The impact of computers
on the government
JAMES A. WARD
Office of the Director of Defense Research and Engineering
Department of Defense, Washington, D. C.

Imagine a large mass of nonhomogeneous metal.
Imagine that it is struck a severe blow by a large
hammer. The shock wave of the impact will travel
throughout the mass and arrive at different parts at
different times. Furthermore, this shock wave will be
reflected from boundaries and regions of nonhomogeneity. These reflected shock waves will affect some portions of the mass time and time again. Occasionally,
shock waves reflected from several directions will converge on some internal region to give an aftershock as
great (or greater than) the original shock. Such shock
waves may persist for a considerable time.
The impact of computers on the federal government
has been analogous to such a hammer blow. There has
not been an impact, but many impacts. These have
struck various government agencies at different times.
In some places the impact has struck again and again.
However, the impact of computers differs in a very
fundamental respect. The shock waves in the metal
attenuate as they convert into heat, whereas the "shock
waves" in government which generate heat seem to
become amplified and increase the heat. Let us look
at a few of these shock waves and the impact they have
had on the federal government.
The first electronic digital computer was the ENIAC.
It was placed in operation at the Moore School of
Engineering just 20 years ago, in the fall of 1945, and
was used for 10 years. It was built by the University
of Pennsylvania for the Army at a cost (including
peripherals and air conditioning) of $486,804.22. As
for the impact, its 30 separate units weighed over 30
tons!! It had 19,000 vacuum tubes, 1,500 relays, hundreds of thousands of resistors, capacitors, and inductors and miles and miles of wire. It used 200 kilowatts
of power, including power for the air conditioners.
The ENIAC was a computer which could add, sub-

tract, multiply, divide, and extract square roots. By
today's standards, however, it had a very limited capability. Its total memory was 20 words (which was not
really a memory in our present sense of the word),
more like hardware holding registers. (However, a
memory of 100 words of core storage was added in
1953.) Programming was somewhat analogous to plugboard wiring, and maintenance problems were severe.
For a time, though, the EN lAC was the best general purpose digital computer in existence. It was
invented to solve trajectory problems using the differential equations of motion, and, as such, it was an
astounding success. A trained operator with a desk
calculator could compute a 60-second trajectory in
20 hours; the ENIAC could do it in 30 seconds, just
half the flight time. It became the major instrument for
the computation of all ballistic tables for the U.S.
Army and Army Air Force. In addition it solved
problems in weather prediction, atomic energy, cosmic
ray studies, thermal ignition, random number generation, wind tunnel design, etc.
Following the ENIAC, as each new computer was
placed in operation it made a considerable impact,
both in the government and the scientific world. The
speed with which it could solve scientific problems
and the number of desk calculations and clerks it could
replace were items that impressed everyone.
The UNIVAC at the Bureau of the Census in 1951
was the first computer applied to business type problems and as such made a tremendous impact. Those
using it said at the time that it was doing a better and
more complete job for half the cost. Speaking of this
computer, a writer in the Scientific Monthly for February 1953 said: "There is in sight a clerical revolution
that could match the Industrial Revolution." Though
this sounded far-fetched at the time, it has turned out
37

The impact of computers on thegovemment

to be prophetic.
Here are some data from the "Inventory of Automatic Data Processing Equipment in the Federal Government 1965," published by the Bureau of the Budget.
There are an estimated 2,188 digital computers in 35
agencies of the federal government. In FY 1966 automatic data processing utilized 63,000 man-years of
personnel and cost $1.1 billion for acquisition and
operation.
What are all these computers doing? Two-thirds of
them are in the Department of Defense. There they do
missile simulation, test data reduction, and war gaming,
and are applied to many kinds of scientific problems;
they do inventory and management work, too.
For instance, one of the largest applications is in the
Air Force Base Supply System now being implemented.
When completed there will be some 150 computers in
a single network using standardized programs and
procedures that will enormously simplify management.
It will reduce the training necessary for operation and
maintenance and is expected to reduce the overall manpower required by some 800 spaces. Most impressive
of all will be the benefits from better efficiency and
faster response times.
The DOD spends almost half a billion dollars a year
on jet fuel. Every six months the Defense Fuel Supply
Center receives and evaluates about 100 bids from
industry to supply the 2 billion gallons required. To
make sure that "the price is right" requires over 20,000
arithmetic computations. To do the task a computer
system uses several different mathematical models.
One model solves the complex bid conditions and takes
care of the import quotas. Another finds the least cost
route from every source to every destination. A third
considers the most logical answers and determines the
minimum cost award pattern, assuring that the problem
is considered as a whole and that the least costly solution is used. Solution of this complex problem formerly
required seven weeks by hand; now it is done in one
or two days.
In the Washington suburbs is an organization of
another type that keeps on magnetic tapes inventory
of all the tactical equipment on each Polaris submarine
by name. This saves a great deal of time and effort
in modifying and supplying them.
A different kind of organization is the Naval Command Systems Support Activity (NAVCOSSACT)
set up' by the Navy to develop and test the programming for the large naval systems. Among these are
operations control centers for the Atlantic, for the
Pacific, and for the Naval Forces Europe. NAVCOSSACT is a large operation with an annual budget of
$14,000,000. It has nearly 300 programmers and

analysts and is trying to hire 50 more.
After the Department of Defense, the world's largest
user of computers is probably the National Aeronautics
and Space Administration. Computers are a vital part
of the NASA system that places humans into orbit,
monitors their position in the space environment and
brings them safely down. These computers receive
radar data, compute orbital parameters, direct communication antennas, and provide handovers from one
tracking site to the next, thus enabling us to observe, on
TV, long continuous periods of communications with
the astronauts. On a typical Gemini flight there are
some 34 computers on-line receiving data in real time
and another 30 or 40 computers working off-line.
About a third of the on-line computers are at Houston,
Goddard, and Cape Kennedy. The rest are at other
sites around the world.
The Mariner flight to Mars is an· astounding achievement-again, vitally dependent upon the use of computers. The navigation and midcourse maneuvers are
two of the most important applications of computers.
The Atomic Energy Commission has received one
of the greatest computer impacts. With extremely large
computation requirements in design, simulation, and
data reduction, the AEC has been a pioneer in largescale computer usage. This application has been told
elsewhere.
In 1964 the Treasury Department issued 370,000,000 checks and bonds. Had. it been necessary to type
them out it would have required 9,000 employees. By
use of computers the job was done by 1,300 people;
besides, the recipients got their checks much faster.
Between 1949 and 1962 the cost per thousand 'for
issuing checks has been cut in half.
The Internal Revenue Service reports a very unusual
impact of computers. Its plans to use computers to
check personal incomes in one of seven of the national
regions were announced to the public; the result was
that in 1964, the public reported $2 billion more interest income than in 1963!
The Bureau of Census used computers in the 1950's
at a considerable saving. The Bureau obtained better
computers and peripherals for work during the 1960's.
Though the amount of work has increased (for example
the publications doubled) it accomplished its work 6
to 12 months earlier-and with 6,600 fewer man-years,
a saving of 25 % .
A catalog of the impacts that computers have made
on each government agency could go on and on. But
there is another kind of impact. The very large activity in computers in 35 government agencies has
caused considerable indirect or secondary impact on
other government agencies.

The impact of computers on the government

For example, the Bureau of the Budget prepares an
annual inventory which lists the government commercial computers. It has published an "Automatic
Data Processing Glossary," and has prepared a report
to the President on the Management of Automatic Data
Processing in the Federal Government. It has developed
procurement procedures that make sure each installation concerned gets a computer properly suited to its
needs and that all qualified suppliers are considered
equitably.
The General Services Administration negotiates with
computer suppliers and publishes price schedules of
equipment. Thus government agencies do not compete
with one another in obtaining computers. GSA is expected to play a larger role soon.
The Civil Service Commission has set up separate
job descriptions for 14,000 Civil Service personnel
now working directly with computers. It had been estimated that the number of government clerical positions would decline with the advent of computers, but
this has not been the case. Though over 5,000 jobs
have been eliminated, more than 7,000 have been
created. The number employed in automation is increasing at a rate of about 1,000 per year.
The National Bureau of Standards provides advice
on computers to all government agencies and will continue to play a leading role in computer research.
Since a number of computers are sold abroad (over
$200 million worth in 1964) they make an impact on
the· gold flow problem and the Department of Commerce gets involved.
The Department of State is involved in the use of
computers too. Under the Mutual Defense Assistance
Control Act of 1951, there is a list of items embargoed
fDr sale (or sold only under restricted conditions) to
the Communist Bloc countries. Computers are on the
list. At present, representatives of State, Commerce, and
Defense are in the process of redefining the U.S. position on computers and other items in this list.
Since 1957 the National Science Foundation has
aided educational institutions in their work with computers. This agency helps colleges and universities with
computer rental and purchases at an annual rate of
$9.5 million. It also finances computer research and
conducts summer institutes. In all, it spends over $10
million per year on computers for others.
This introduces computer education, an important
topic in itself. For reasons to be brought out later,
computer education (or training) is in relatively "short
supply" at all levels in the Department of Defense.
Last year special courses were given to 30,000 people.
In addition, 20Q were sent to colleges and universities
for graduate work in computers. These courses included

programming, hardware design, systems design, numerical analysis, etc.
Each of the military Academies, Army, Navy, Air
Force, and Coast Guard, has one or more computers
for use of the faculty and cadets. As of this year, each
entering cadet is required to take a course in computer .
programming. Almost all of them are required to use
computer programming in solving problems in their
other courses. Each Academy also encourages the
cadets to use computers on their own, and many do.
A great many senior officers and civilians in the
Department of Defense have to work with .computers.
However, computers were invented after our present
senior officers and civilians finished their formal schooling. To assist them, there was set up, last year, the
Department of Defense Computer Institute (DODCI).
At present it gives two courses: a I-week (50-hour)
course for generals, admirals and supergrade civilians
and a 2-week (80-hour) course for Colonels, Naval
Captains and civilians of comparable rank. (An additional course is planned to begin next year.) The idea
is not to make these senior managers experts, but to
give them some of the concepts, some of the language,
and some notion of what it takes to make a computer
system (especially a military system) work. The courses
have been enthusiastically received.
Interest in computers has also extended to the Congress. Many hearings have been conducted, and during the last session the Brooks Bill was passed. Among
other things, this Bill arranged for a computer administrator in GSA for the management of computers
in the federal government and for a pool of funds
for procurement. Just how this arrangement will interact
with the present arrangements has not yet been worked
out. Perhaps this aspect would make an interesting
topic at a future meeting of AFIPS.
Thus we see that, in addition to those who work
directly with computers, there are many others-even
entire agenCies-that feel the impact of computers on
the government.
As was stated earlier, the Bureau of the Budget inventory lists some 2,200 computers in the federal government at an annual cost of over $1.1 billion including
manpower and operation cost. Everyone of them is
needed. During the current fiscal year some 260 more
are expected to be added, and the total cost including
operation is expected to rise slightly. Besides these,
contractors are estimated to utilize about another 1,800
computers on government contracts.
However, there are other entirely different and additional types of computers that are perhaps even
more important to the Department of Defense-the
military computers utilized as integral parts of weapon

The impact of computers on the government

systems. These come in various sizes, shapes, and
capabilities. and some of them perhaps can only be
classified as computers in the broadest terms.
It is difficult to determine the total number of U.S.
military digital computers. An informal survey of our
manufacturers show that over 12,000 have been delivered with 1,800 of these delivered this calendar year.
It may be of interest to note that three manufacturers
have delivered over 1,000 military digital computers
each. They are, in alphabetical order, Hughes Aircraft,
North American Aviation, and Raytheon. One of them
had a special news release a. few months ago on the
occasion of the delivery of its 3,000th computer.
A vital part of our national defense is our missile
capability, and therefore this defense depends on the
computers that these missiles utilize. Such computers
are different in many respects. For example, they are
programmed by the manufacturer and management is
not worried about how much they are utilized!
For instance, most types of missiles have inertial
platforms that provide data to special purpose digital
computers which perform the missile navigation computation. Such a computer usually includes a digital
differential analyzer (DDA) "with a wired-in program.
However, it has a memory, can add and subtract, caif.
make decisions, ~accept data and give output.
The first missiles were checked out by hand, but with
the vast number of electrical levels that must be set and
verified, it soon became evident that a computer could
do this much more quickly and with greater accuracy.
In Civil War days the cannoneer sighted down the
cannon barrel, estimated elevation, and fired. To aim
missiles at targets hundreds of miles away is an entirely different matter; this is a task for a computer
which determines the proper parameters and enters
them into the guidance computer contained in the
missile.
This "aiming" computer on a POLARIS submarine
is a very respectable computer. The POLARIS submarine also has computers for navigation. The Ships
Inertial Navigation System (SINS) contains a DDA
to integrate the output of the accelerometers. The SINS
is connected to a general-purpose navigation computer.
All these computers are replicated for reliability. Computers have made quite an impact on the "submarine
business." Instead of being restricted to attacking
targets that they can "see," the computers enable submarines to act as a strategic weapons platform that
can be used to attack targets over 1,000 miles away.
Digital computers in military aircraft aid the pilot
in navigation, receive radar data, calculate intercept
courses, assist in the aiming of missiles, in dropping
bombs, etc. In fact, inertial navigators with digital

computers are in test phase on several large commercial
jet planes.
Radars used for missile and satellite tracking have
special-purpose digital computers. Some also have
general-purpose computers for converting orbital parameters into look angles, and vice versa.
Another interesting application of computers is the
Department of Defense communication network AUTODIN (Automatic Digital Information), a nationwide
network for the transmission of digital messages. There
are 6 "switches" with up to 250 subscribers per switch;
each switch contains a computer with considerable
storage capability. From the header of an incoming
message the computer determines where the message
should go and its priority. As soon as its priority warrarits, the message is forwarded to the next switch
and/ or to its destination. It is an all-automatic network.
Then there are many kinds and sizes of military
general purpose computers. Most of them were designed by R&D contracts and especially suited for a
particular task. The ANFSQ-7 used in the SAGE
centers is an early example. It is very reliable, and
has the capability to handle an unusually large number
of inputs and outputs.
Most general purpose military digital computers are
built to withstand a physical environment of temperature extremes, vibration, and shock that could not be
tolerated by the average commercial computer which
operates very well in a stable air conditioned room.
They are also required to pass more stringent reliability
tests.
When John Paul Jones stood on the quarter .deck
of the Bon Homme Richard he could see and evaluate
every threat to his own ship and to any in his task
force. Today a task force covers hundreds of square
miles with picket planes out beyond those boundaries.
Their combined radars cover thousands of square
miles. By means of NTDS (Naval Tactical Data
System) all the targets of all these radars are processed
by computers, transmitted, correlated and displayed on
a scope for the task force commander. With the aid of
computers (and radar and communications) the commander can "see" his situation as well as could John
Paul Jones. To date the system is on 10 ships and has
cost somewhat more than $300 million, including
RDT&E. There are two to three computers per
system and it is truly a mUlticomputer system. Other
multicomputer systems that I know of have one
computer operating and another on standby; the NTDS
system has the job split up among the computers.
The Army has the well-known truck-mounted
BASIPAC and MOBIDIC. In addition, it is in process

The impact of computers on the government

of fielding a substantial number of F ADAC computers.
These are computers designed to operate in the field
to compute firing data for an artillery battery. N apoleol\
once said that God fights on the side with the "most
cannon." The F ADAC will add so greatly to the surprise ("first round accuracy") and effectiveness of the
artillery, that it will have the effect of being the "most
cannon."
No account of the impact of computers on the government or national computer community would be
complete without mentioning SAGE, i.e., the Air Force
Semi-Automatic Ground Environment system. Undoubtedly, a nontrivial portion of this audience has
been directly affected by it and many others have been
indirectly affected.
SAGE is an air defense system whereby radar data
is digitized, transmitted to computerized centers, processed, displayed and, in some cases, transmitted to
other centers. It involves a great many technologies,
but let us look only at the computer aspects.
Some people measure impact in dollars: the computers for SAGE cost $1 billion, and the programming
an additional $110 million.
There are 150 special-purpose computers to convert the radar data to digital form for transmission to
the computerized centers. In each center there are two
(for reliability) large special general-purpose computers.
The first installation of the system was placed in
operation at McGuire AFB in July, 1958. SAGE built
up to 25 centers, and at present 19 remain in operation.
The programming for SAGE was a very large task.
The cost of programming Modell was $18.5 million.
As more capabilities have been added the programs
have been repeatedly revised until today Model 9 is in
operation. The programming contract for the current
fiscal year for modifications and improvements of
Model 9 is budgeted at $8.4 million. The number of
instructions in the present system is over 400,000 with
about 100,000 at each installation. In addition, the
active utility programs contain about 750,000 instructions.
The number of programmers required was originally
estimated to be about 600, which at that time (1951)
was about the total in the U.S. Hence, SDC (the programming contractor) has had to train (partly as a
result of personnel turnover) some 2,000 programmers.
This is not the time, place, nor the speaker to comment on the operational achievement of the entire
SAGE system. However, the computer systems therein
have made quite an impact. Some important points
are:

1. The ability to process radar data in real time.
2. A man-machine interaction through data display and manual input, bringing the computer
capability to engineers, managers, and decision
makers.
3. The extensive use of computer-to-computer
communications over a distance of miles.
4. The feasibility of over 100 people time-sharing
a computer on a real-time basis by means of a
device like a satellite computer.
5. The ability to switch automatically over to the
standby system.
6. The development of sophisticated software utility systems.
7. The COMPOOL concept of many subprograms
sharing and processing data for each other.
Another class of general purpose computers is used
in Command and Control Centers. There are many
others such as general-purpose computers designed to
operate in airplanes, in vans, in helihuts.. There are
also a number of one. . of-a-kind computers and those
designed to be used as training devices.
There are "odd ball" computers in the government
other than military. NASA has put especially designed
computers in satellites. Others· are used in launching
and tracking stations.
Process control computers are also used in the government. Most of those sold now are general-purpose
computers with special I/O capabilities and designed
for great reliability and rugged environment such as
vibration. They not only control machine tools but
power distribution networks and power generating
facilities.
The TVA has 200 power generating units with
different kinds of fuel, different efficiencies, and differing costs of operation and stream flows. It has 1.5
million customers spread over 80,000 square miles.
How to generate and distribute the power in the most
efficient manner when loads are changing by the
second is a task that computers do for TVA.
Let us now consider computer research. A little over
a year ago, in an article in Datamation, a well-known
computer user had this to say:
Five years ago, the Government should have init\ated
research into the problems of constructing digital computers 100 to 1,000 times as fast and 1,000 to 1,000,000 times
as large as far as rapid access storage is concerned. It is
absolutely absurd in a world in which science and technology are obviously of major importance to allow the
development of this fundamental scientific tool, the digital computer, to be left to the whims, chances and tax
write-off of commercial computer companies.

Of course, our government continues its very large

The impact of computers on the government

support of computer research and development-but,
understandably, in ways that are expected to result
primarily in the solution of its own pressing problems,
especially in national defense. A rough unofficial estimate of Department of Defense annual expenditure
on computer research, development, test, and evaluation (including hardware, software, peripherals, and
weapons systems) is $280 million. The present day
development of computers has resulted, to a considerable extent, from the continued cooperation between
government and industry beginning with the Army's
original contract to build the ENIAC.
There could be several reasons why the major
government support of research on computers is perhaps not as generally well known as it might be. Most
of the work is done (and properly so) by the computer
manufacturers and even though funded by the government, is reported upon and administered by those
individual industrial concerns. Only a relatively small
portion is done in the government laboratories, though
most of it is monitored by and funded through them.
Also, computer R&D cost is often lumped in with the
weapon system in which the computer is used, and
separating it is like extracting a pound of flesh without
loss of blood. A third reason is that the payment for
a contractor's R&D may be indirect. For example,
government procurement contracts allow "overhead"
and, under certain conditions, this overhead can be
used to support a company's "own" IR&D (Internal
Research and Development). Just which company
tasks are government-funded is settled by negotiations
and does not necessarily appear explicitly in any contract. The amount so spent is over $1 million per year
by each of a number of companies.
The computers in each of the weapons systems have
gone through RDT&E. Such costs for NTDS (Navy
Tactical Data System) have been over $95 million.
The Marines are developing their second tactical data
system. The Army is developing a computer for the
new NIKE system. The Army also expects to spend
$10 million this year on their ADSAF program (Automatic Data System for the Army in the Field). For
this they are planning to use van-mounted computers
for the troops in operations. They are planning a supply
system, a tactical fire planning system, and a tactical
operation system for the commanders of field armies.
Napoleon could stand on a hill and see his and the
opposing forces and their movements. A modern army
covers thousands of square miles. The Tactical Operation System is a realtime, on-line, computerized, communication and display system that will give the commander an up-to-the-minute view of the tactical situation.

Besides the many large systems presently being developed throughout DOD there are many hardware development programs: associative memories, cryogenic
computers, mass memories, complete new computers,
physically small memories, sub-nanosecond circuitry,
and various peripheral devices including displays.
It would seem that in a tactical environment (airplane, ship, truck) a mass memory with no moving
parts would have an advantage over a drum or disk.
Such a device should be cheaper than core (even if the
core is fabricated in Hong Kong), small, and rugged.
So far 13 techniques have been investigated and several are being funded. Some of these methods are (1) a
plating and etching process that gives the effect of wirestrung cores, (2) thin-film plated wire, (3) wire screen
woven on a special loom, (4) a "delay line" that remains frozen but can be rapidly shifted as desired,
( 5) moving domain walls, (6) the residual magnetism
left in a metallic rod where an electrical pulse overtakes
an acoustical pulse, and (7) artificial crystals.
Also, entirely new types of computers are being developed. For instance, the Air Force is building one
without conventional commands. The programmer
specifies in a separate memory what his commands are
and then writes a conventional program in the conventional way. Such a computer will be able to operate
like-not simulate but operate like-any of several different existing computers and accept their programs.
It is also a tool that can be used to determine what
hardware organization and command repertoire is. optimal for a given job or with a given compiler language.
The first computers were vacuum tube computers.
In the transition to transistors the immediate achievement was the reduction, by a factor of 10 or more, in
the physical size, power requirement and air-conditioning requirement. However, the real impact has been in
reliability increase and cost reduction.
In the tube computer days the maintenance· man arrived before daylight to check out the computer before
the working day began, and he was in attendance as
long as the computer was operating. Early reports on
the first UNIVAC at the Bureau of the Census show
that they wished to run the computer 24 hours a day,
7 days a week. However, during some weeks the computer was down 60 % of the time. Let us be generous
and say that the average vacuum tube computer had
a mean-time-before-failure (MTBF) of 10 to 15 hours.
In the early days of transistors a contract request for
delivery of a new military computer was almost turned
down by one manufacturer because of the reliability
requirement: it was required to have an MTBF of 50
hours. That would be almost like asking for a 50-year
MTBF today. Now many military computers have well

The impact of computers on the government

over 1,000 hours MTBF and there is one case of a
military computer that ran over 10,000 hours without
a failure. The reliability know-how developed for military computers has been applied to the design of commercial computers so that they also can be quite reliable now.
Along with this increase in reliability has come a
reduction in cost. For instance, one particular computation problem that cost $14.40 in 1948 is expected to
cost 2¢ in 1966. Reduction in cost and increase in
reliability have been probably the major reasons for
the rapid expansion of computer use in the government
and in industry. This pattern of reduced cost and increased reliability is expected to continue in the third
generation of computers-that using integrated circuits.
For some time now the Department of Defense has
been sponsoring research in integrated circuits. As usual,
the first application of this technology has been in military computers. The result has been a remarkable reduction in size, weight, and power. A dozen or more
companies are now in the process of building such
computers, and several are being shown at conventions.
One such operating computer is the guidance computer
in the Minuteman II: it is 6.3 X 10.3 X 21.3 inches,
including memory and power supply. However the
most remarkable aspect of integrated circuit computers
is the increase in reliability. Its computed MTBF is
measured in years.
Compare this performance, if you will, with that now
considered acceptable commercially. I know of one
computer, on two-shift operation, that is "down" about
four hours a month; the manager feels this is normal.
In advancing from tube to transistorized computers,
the MTBF went up by a factor of 10. At least another
factor of 10 increase is expected going from discrete
to integrated circuits. In a short time, hopefully, this
technology will be cheap enough to be used in commercial computers. Some companies have announced computers using such circuits.
Already coming over the horizon is what looks like
the fourth generation of computers: MOS (Metal Oxide
Semiconductors). Their use promises to yield yet another major reduction in size and increase in reliability.
Already, manufacturers are seriously considering using this technique to put the entire arithmetic unit of a
computer on a single 3 X 4 inch logic card, or a complete DDA (except for memory) for a missile guidance
computer on a single chip the size of a quarter. With
some of the new miniaturized memories it seems likely
that a few years hence a computer with. the capability
of the most common ones now in use could be carried
in a briefcase-or even attached to the underside of
the input typewriter!

The speed with which the technology is advancing is
aptly illustrated by a recent note in the New Yorker
magazine. The note consists of two sentences. One sentence quoted from the text of a new book: "While you
are reading this sentence, an electronic computer is
performing 3 million mathematical operations!" The
author probably had in mind a well-known commercial
computer that is advertised to perform 3 million operations per second. But before the book could be published, the author and publisher heard of newer developments so that the dust jacket has the following
sentence: "While you are reading this sentence an
electronic computer is performing 4 billion mathematical operations!"
Considerable research is also being done in software.
This includes language translation, compiler development, information retrieval, pattern recognition, operational systems, and time-sharing systems, to name but
a few.
However, it is now quite clear that a great deal of
further research is needed to make computers more
generally useful. The original computers were designed
to solve differential equations and other mathematical
problems. The design was extended to make business
type applications easier. They were designed for batch
processing and to be programmed in machine or symbolic language.
Now computers of this design are being used for
nonnumerical operations such as command and control, information storage and retrieval, language translation, etc., and are programmed by compiler. What
logical hardware organization is needed for this kind
of work?
More efficient compilers are needed to make existing machines operate more efficiently. A language with
5-character labels will not work very efficiently on a
computer with 4-character words!
Computers are designed and then compilers are
written for them. In my view this process should be
completely reversed. If a computer is intended to be
used mostly through a higher-order language and compiler, it should be designed to do this easily. In fact the
language-compiler designer, the executive system designer, and computer hardware designer should cooperate on an integrated hardware-software package.
The designers of the hardware and software for the
new computer for Nike X have so cooperated. Though
the forthcoming computer is faster than the preceding
ones, the throughput increase is much greater than the
hardware speed increase.
Also, better peripheral equipment is needed, especially for tactical military systems. Better kinds of
equipment and programming systems for man-machine

The impact of computers on the government

interface is needed. More people need to be educated
on how computers work and what their limitations and
capabilities are.
In summary, then, we have seen that computers
have made many kinds of impacts on the government.
An annual cost of $1.1 billion and 63,000 man-years
of effort is no mean impact. By taking over repetitive
tasks the computers get the work out faster, more accurately, and with fewer people-as illustrated earlier
by the Treasury Department check-writing example.
Saving substantial amounts of money and time always
produces an impact. Computers enable tasks to be
undertaken that would otherwise be impossible such as
the NASA manned space flights. The utilization of
computers has caused indirect impact on other government agencies, such as the Civil Service Commission,
the Bureau of the Budget, the National Science Foundation, etc.
The Department of Defense has received exceedingly
strong impacts. Not only does it have two-thirds of the
commercial-type of computers in the Federal Government, it also has many and diverse types of military
computers; many weapons systems have computers as

integral parts thereof; computers are also essential parts
of many tactical and strategic systems.
The Department of Defense (and to a lesser extent
other government agencies) is spending a very considerable effort on computer research across the board
(hardware, software, peripherals, and systems). As
they become available, the fruits of this research will
cause impacts that will continue to reverberate throughout the government and in industrial and commercial
enterprises as well.
It is told that someone asked the late Norbert Wiener
if the government could now do away with its com
puters. He is said to have replied that the government,
having had computers, could not possibly go back to
doing without them: that it was like Adam and Eve
having eaten the forbidden fruit-they could never be
the same as before.
I would like to add that eating the forbidden fruit
gave Adam and Eve the ability to tell right from
wrong and the incentive to acquire wisdom. I trust that
all of us will use our computer-aided wisdom to make
the sensible decisions for better utilization of the computers we have and to invent better ones.
4

Communications, computers and people*
PAUL BARAN

The RAND Corporation
Santa Monica, California

terminals. As we expect to be talking about digital uses
which by definition are digital terminals, we shall confine our observation to telephone switching and transmission.
At present, the telephone plant, our prime data carrier, is almost exclusively based upon electromechanical
switching-that most primitive form of computer logic
-and one that we in the computer business haven't
seen around for years. Transmission is by means of
frequency division multiplexing--or about as analog
(or undigital) an operation as we computer types can
envision. The only kind words a computer man can
have for this system is that it works; it works well and
has been working well for many years-for the purpose
for which it was designed.
While perhaps slow by pace, electronic switching
has arrived on the scene for the telephone company.
At least two separate systems in this country have
now passed field trials and are being installed commercially. This new equipment may be representative of
the future telephone local central offices. At present,
these electronic switches are not believed to be more
economical than their earlier electromechanical switch
counterparts. But their prime advantage lies in the
new additional services that they offer because of the
general computer nature of the control mechanism of
the switching center. For example, it will be possible
to dial only two digits to reach the few numbers that
you call often. It will be possible to relay a call to
another telephone if you are temporarily away. Automatic diagnostic routines will permit repair and mainte-

INTRODUCTION
Communications and computers are today becoming
what the economists call "complementary goods"--one
without the other is of much- lesser value-like pen
and ink, pretzels and beer, and gin and dry vermouth.
Let us first briefly consider the impact of the computer technology upon the communications business
and, conversely, how good, widespread, low-cost digital
communications will allow a dramatic increase in the
creation of new types of computer systems. Then we
shall get down to the meat of the talk-a few of the
unappreciated social consequences possible and, lastly,
we shall proffer remedies in advance of the time society realizes there is a problem. If the order of things
appears backward, with remedies being offered in advance of the patient's complaining of an ailment, it is
due to our belief that the lead time for the cure of social ills is often longer than the gestation period of the
disease. Only we who appreciate what is happening to
computer development may be in the best position to
see the thunder clouds.
THE IMPACT OF COMPUTER
TECHNOLOGY ON COMMUNICATIONS
Communications equipment is sometimes categorized
into switching equipment, transmission equipment, or

*Any views expressed in this paper are those of the author.
They should not be interpreted as reflecting the views of The
RAND Corporation or the official opinion of the policy of any
of its governmental or private research sponsors.
4S

Communications computers and people

nance by inexperienced personnel. Further, electronic
switching is a new technology whose price is expected
to decline rapidly in the future.
So much for switching. We also see computer technology creeping into the picture in transmission. The
Bell System T -1 multiplexing system samples 24 analog
voice channels about 8,000 times per second producing,
together with synchronizing information, a data stream
of 1.54 megabits per second which can be transmitted
over ordinary copper pairs using pulse regenerative
amplifiers. This pulse code modulation technique is
being developed simultaneously in many countries and
is in use but presently restricted for links on the order
of magnitude of about 20 miles.
As pulse code modulation is the most economical of
the multiplexing systems, it appears destined as the
transmission direction of the future. Even though digital
technology is entering the telephone plant slowly and
in a piecemeal fashion, it is arriving and will make its
impact felt. Specifically, most of the growth of the
telephone plant may be expected to occur within these
digital techniques areas. The implications to us are
severalfold.
First, as we expect to see a rapid drop in the cost
of digital circuits, we may expect continued drops in
the price of digital communications in the future. We
. would also expect to see even more marked savings to
the digital communicator as systems evolve which are
more amenable to the all-digital processing of information from user to user. Today emphasis must be given
to complete compatability with the large existing analog
system in being. where periodic reconversion to analog
signals is required. Thus, one day we envision the bulk
of the telephone system being built entirely of digital
processing assemblies in lieu of the all-analog systems
as of today. When this day comes, we computer types
would view the telephone system as merely another
particular computer application and not necessarily a
specialty field unto itself. If the old-time telephone engineer fears that the computer types are taking over, he
is probably right. So much for what computer technology might do for or to communications, depending on
where you sit.
THE IMPACT OF COMMUNICATIONS
UPON COMPUTERS
Using telephone lines modified to handle digital data,
we are able to build an increasing number of geographically distributed time-shared computer systems. Many
individual users are connected to a common computer
data base. Examples of such systems include airline
reservation systems for civilians and fancy display "com-

mand and control" systems for the military.
Simple record keeping, a mark of a highly developed
economy, has been a prime area of development of
these large computer file/ communications systems
where much .of the routine clerical work is transferred
to the computer with human interrogation of the system. As time moves on, the number of people who
will be able to interrogate the system and the geographical distance between them and the machine will increase.
SOME INDIVIDUALLY USEFUL SYSTEMS
Today we see time-shared file systems used for insurance records, for checking automobile tags, to locate
outstanding criminal warrants, and for credit check
investigations (using drivers' license numbers) in cashing checks. The systems built to date pose no overt
social problem. The information handled is not highly
sensitive and access to it is generally limited.
THE TRAIL OF ARTIFACTS
IN A CIVILIZED LIFE
As we pass through life, we leave a trail of records
widely dispersed and generally inaccessible except with
a _great deal of effort and diligence.
, We start with a duly recorded birth certificate. We
leave behind hospital records and our pediatrician adds
to our medical records. We are deductions on our
parents' income tax. School is a place where we busily
generate record upon record of our scholastic grades,
our attendance, our IQ test records, our personality
profile records, volumes galore. With automated teaching coming to the fore, we can expect better record
keeping. The volume of data we will record per child
may be expected to increase even more markedly ("in
the best interests of the student"). Between terms we
get our social security card and a job, and we start
leaving behind us a long history of employment records.
We reach age 18 and are entered upon the records of
the Selective Service. We get a driver's license and, if
we are lucky, we will be able to avoid having arrest
and jail records. Most of us will apply for a marriage
license, some of us will collect divorce decrees which
will end in voluminous court records. We move from
job to job in a mobile economy creating movingcompany inventory records of our goods. Even as we
move from place to place we leave behind short records of our airplane reservations and for some reason
every hotel makes a ritual of acquiring and preserving
the names and addresses of its guests for posterity.
This list is only a partial one. Play the game yourself

Communications computers and people

and think of all the records you leave as you go
through life.

information available from a portion of these potential
records. The price for a fishing expedition for information is high and most of the fish are inaccessible.

WHY SO MANY RECORDS?
One does not create records merely for the sake of
creating records. But rather there is the implicit assumption that the records will be of some use some
day. In order to be of use, there must be some means
of interrogating the files to resurrect the information
sought. Thus, we envision large families of systems,
each individually useful. For example, an Internal
Revenue Department investigator might wish to have
immediate access to the tax returns of the associates of
a man who is being audited to check for consistency
of financial relationships.
A company may wish to have rapid access to its
personnel files to know whether to give a good reference
to a former employee.
A doctor may wish to trace the entire medical history of a patient to provide better input into a diagnostic
computer.
The Veterans Administration may wish to examine a
man's complete military record and possible other
previous medical records to see whether the ailment
claimed as being service-connected really is.
A lawyer for the defense of a man will wish to
search for jail records, arrest records, and possibly
credit records of all witnesses for the plaintiff.
Professional licensing boards may wish to delve into
any records that may indicate the applicant lacks an
unblemished character.
The military in filling extremely sensitive positions
may even wish a record of all books borrowed by the
. prospective applicant to insure that his interests are
wholesome and he possesses the proper political bias
desired.
ACCESS TO THIS INFORMATION
Today one does not gather such information about
the prospective examinee easily. If one went through
the direct channels and asked most sources for their
records about a person, he would most likely be told to
go jump in a lake, if for no other reason than the information is not available--cheaply. Even if the information were a publicly available record, the investigator must be expected to spend a great deal of time
and effort delving to discover pertinent data. Today, as
through a practical matter, if one wishes to obtain
much of this information about a person, he hires a
private detective who charges ,a great deal of money
and expends a great amount of time obtaining a little

THE IMPENDING PROBLEM .
So much for the pleasant past. Consider the following argument:
1. A multiplicity of large remote-access computer
systems, if interconnected, can pose the danger of loss
of the individual's right to privacy-as we know it
today.
2. The composite information data base may be so
large and so easily accessible that it would permit unscrupulous individuals to use this information for unlawful means.
3. Modern organized crime should be expected to
have the financial resources and access to the skills
necessary to acquire and misuse the information in some
of the systems now being considered.
4. We are concerned not only with the creation of
simple "automated blackmail machines" using this information, but with the added implication of the new
"inferential relational retrieval" techniques now being
developed. Such techniques, when fully refined, could
draw chains of relationships from any person, organization, event, etc., to any other person, organization, or
event.
S. Humans, by their day-to-day necessity of making
decisions using totally inadequate evidence, are innately
prone to jump to conclusions when presented with very
thin chains of inferred relationships. For example, merely plastering a man's name on billboards will markedly
change the outcome of an election, if the other candidate's name is not equally displayed.
6. The use of private detectives to unearth derogatory
information on political candidates and their associates
has become an increasingly prevalent feature of elections. This practice is expected to increase in the
future.
7. The cost-per-unit-dirt mined by unautomated human garbage collectors can be cut by orders of magnitude once they obtain access to a set of wide-access
information systems which we now see being developed.
It is the sophisticated form of chain-relation blackmail
that may be of most social concern. We generally pass
through three stages of information storage development. First, we start by keeping manual records employing clerks. Next, we get rid of some of the clerks
when we put all the records into a single central computer file with the readout controlled from a single
point. The next step is the creation of remote interroga-

Communications computers and people

tion devices to interact with the file from a large number of points. The payoff for instant access is often
high as it eliminates all delay to the file user.
8. This development of geographically widespread
access systems requires the use of communications lines
to connect the users into the computer. There is a
widespread belief that somehow the communications
network used will possess a God-given sanctuary to
privacy, but "it ain't necessarily so ... "
DIRECTIONS TOWARD A SOLUTION
1. Assume that not everyone is as honest and as
trustworthy as ourselves-but is just as diabolically
clever.
2. Appreciate that we will be increasingly dealing
with complex and, hence, difficult-to-understand-allthe-details types of systems in the future.
3. Probably the only people who best understand
the operation of each system will be computer design
engineers who build the system in the first place.
4. Often the only time that the fundamental safeguards that we seek can be applied is at the time of the
initial system design. "Software patch-ups" at a later
date may generally be relatively ineffectual compared to
good initial design-good design being defined as including an awareness of the existence and importance
of the problem.
5. Do not expect help from the legal profession in
lieu of good design. Even ignoring the social lag of
the legislative/judicial procedure, the detailed subject
matter verges on or beyond the limits of their comprehension. *
6. Laws and laws alone have been almost totally
ineffectual in the growth of widespread electronic
eavesdropping and wiretapping. At most, all the courts
have accomplished is to prevent the police from using
the same techniques available to the private detective
or the criminal-or even casual readers of an electronics technician magazine.
7. While I have little faith that laws in themselves
will solve the problem, laws could be helpful in two
ways: (a) Laws outlawing certain practices will be of
minor help in increasing the price of the act and making
* Eight months after this talk was presented, a special Subcommittee of the Committee on Government Operations
headed by Representative Cornelius E. Gallagher looked into
this problem. (These hearings, entitled "The Computer and
the Invasion of Privacy", by a Subcommittee of the Commitee on Government Operations House of Representatives,
Eighty-ninth Congress, Second Session, July 26, 27 and 28,
1966 are available from the U.S. Government Printing Office
for $0.75.) Because of these hearings and the resulting
interest and action, many of these words are now obsolete.

its commission less flagrant; and (b) laws can be written
so that potentially weak systems cannot be built unless
adequate safeguards are incorporated throughout for
the protection of the information stored.
8 .. This last direction is to me viscerally unsatisfying
as it carries with it a built-in loss of freedom. The
thought of the creation of another governmental agency
peering over one's shoulder contains the seeds of the
possibility of bureaucratic decay and arbitrary conclusions based upon an incomplete understanding of
complex problems.
9. Historically, government regulatory agencies start
as highly effective bodies but lose momentum as the
original personnel leave and their replacements come
from the industry being regulated. (Where else are you
going to get competent people who know the business?)
The extreme competence that we need in a regulatory
agency of this type is too rare a commodity.
10. If we are to avoid external regulation, then it
behooves us computer-communication system designers
to start working, or at least thinking, about the problem. We should take the initiative and the responsibility
of building-in the needed safeguards ourselves before
Big Brother is forced to do it himself and we are not
too happy with the way he might want to do it.
11. Safeguards, whether they be screens around moving machinery or circuit breakers, cost money. Every
design engineer is reluctant to add anything that costs
money and buys little visible protection. But the writer
believes that it is time to start regarding such added
costs as necessary costs-a price to society for the
privilege of building a potentially dangerous system.
12. This is not a new concept. We have, for example,
been practicing this in the design of sewage systems
and in electrical distribution systems for some time.
But, historically, it has generally taken an epidemic to
build a local sewage disposal system. It took a series
of disastrous fires to get our electrical codes.
13. The national geographical extent of the new
data systems, their impact, and their investment are so
large that the price of a "retrofit" after the calamities
occur may be a higher price than we need have paid
if we had used some preplanning.
PROPOSED SPECIFIC SAFEGUARDS
To be more specific, what safeguards do I envision?
Of course, we don't know all the answers yet. But,
clearly, there are several steps that we should be considering, including:
1. Provision for minimal cryptographic type protection to all communications lines that carry potentially

Communications computers and people
embarrassing data-not super-duper unbreakable cryptography-just some minimal, reversible, logical operations upon the data stream to make the eavesdropper's
job so difficult that it isn't worth his time. The future
holds the promise of such low-cost computer logic, so
this may not be as expensive as it sounds.
2. N ever store file data in the complete "clear."
Perform some simple (but key controllable) operation
on the data so that a simple access to storage will not
dump stored data out into the clear.
3. Make random external auditing of file operating
programs a standard practice to insure that no programmer has intentionally or inadvertently slipped in a
"secret door" to permit a remote point access information to which he is not entitled by sending in a "password."
4. When the day comes when individual file systems
are interconnected, let us have studied the problem
sufficiently so that we can create sensible, precise
ground rules on cross-system interrogation access.
5. Provide mechanisms to detect abnormal informational requests. That is, if a particular file is receiving
an excessive number of inquiries or there is an unusual
number of cross-file inquiries coming from one source,
flag the request to a human operator.
6. Build in provisions to record the source of requests for information interrogations.
7. Audit information requests and inform authorities
of suspected misuse of the system.
This list is open-ended, and it is hoped that more
suggestions will be forthcoming. But, to serve as an
example of the need for and type of safeguards we are
talking about, to illustrate how such thinking can
ameliorate the problem of loss of privacy, consider
what we might do in the case of our present telephone
system.
ONE EXAMPLE OF INCLUSION
OF PROTECTIVE MEASURES:
THE TELEPHONE SYSTEM
Today we are deluged with bogus telephone advertising, crank calls, bomb threats, false fire and police
alarms. Obscene telephone calls, particularly to single
women, have become so prevalent that it has been
publicly suggested that female names be listed only
as initials.
In Washington, the number of these calls has become so great that after much Congressional and press
discussion, the penalty for making obscene calls was
raised from $10 to $500. Of course, it is a rare event

when a person making an obscene telephone call is
caught, so the deterrent effect is almost nil. But an
increased penalty hidden in a law book is the standard
legal response to a basically technological! social problem. This writer would prefer to see technology which
created this problem be required to provide more
effective safeguards.
For example, each telephone (or at least those
plagued with these calls) should have a button which
when pressed bridges the call to a bank of recorders
at the police station and a teletypewriter message with
the name, address, and telephone number of the calling
party transmitted to the nearest police car. It wouldn't
take long to clean up the undesired callers.
If you were to make this suggestion today you would
be told that this is not practical because it would be
prohibitively expensive since this requirement did not
exist when the original electromechanical telephone
system was set up. This is true, but let us look at the
emerging use of the all-electronic switching centers
we have been talking about. It will be relatively easy
now to add such an immediate track-back feature. Will
we do it? I don't know. It would cost money and there
are many reasons telephone companies would wish to
avoid getting involved-but here is a perfect example
of the social implication of the instrument which can
violate our right to be left alone. The telephone can
be designed (at a somewhat higher cost) to provide
safeguards forming added protection to prevent it
from being socially misused.
Clearly here is an example of the trade-off between
dollars and the type of society we want. It will fall to
the computer system engineers to make such decisions
more and more often in the future.
What a wonderful opportunity awaits the computer
engineer to exercise a new form of social responsibility.
The advent of the new computer-communications technology need not be feared with 'trepidation as we approach 1984. Rather, we have in our power a force
which, if properly tamed, can aid, not hinder, raising
our personal right of privacy.
If we fail to exercise this unsought power that we
computer engineers alone hold, the word "people" may
become less a description of individual human beings
livmg in an open society and more a mere collective
noun.
It may seem a paradox, but an open society dictates
a right-to-privacy among its members, and we will have
thrust upon us much of the responsibility of preserving
this right.

The impact of computers on urban transportation
KENNETH J. SCHLAGER

Southeastern Wisconsin Regional Planning Commission
Milwaukee Wisconsin

Dr. Ramo has said that one of the purposes of this
series of talks was to try to "close the gap" between
the information technologists and the applications of
their technology in the real world. If this is one of our
primary objectives, I don't believe there is a field in
which the gap between the technologist and the application is so great as in the area of urban planning,
of which transportation is a part. There's definitely a
gap between the person who knows the problem and
the person who knows how to solve the problem. The
reasons for this gap are historical and complicated.
There's no need to go into them at this time, but the
gap, I think you'll discover from this talk, very definitely exists and needs to be" closed.
When one considers the concept of the impact of
computers on urban transportation, he is almost forced
to remark, "Impact, what impact? Change, what
change?" Looking around today in urban areas, you
see crowded freeways, decrepit, decaying transit systems. It somehow all seems the same except that it's
getting worse. The automobile is certainly not a product of the computer. Our old rail and bus transit
systems are certainly not a product of the computer.
So why should we blame it on the computer? Just what
is the impact? What has it been? What is it going to be?
At the other extreme, the Sunday supplement articles recite the wonderful hopes of the future: automatic highways, completely controlled freight terminals,
exotic transit systems. What is the real truth? What
kind of impact has there been and what type of impact
seems to be coming?

The true situation is somewhat between these two
extreme views. The first indications of an impact are
now apparent. More precisely, a dual impact of computers on urban transportation has become evident
in the last few years. The first impact relates to the use
of computers in the planning and design of transportation systems. And inevitably this design application
results in an involvement with the whole problem of
urban design because it is impossible to. separate the
transportation or circulation system in a city or a
metropolitan area from the spatial pattern of the city
itself. The second impact is more direct and involves
the use of computers as part of the transportation
system itself. Today in transit systems, and in the future even in individual vehicle-type transportation systems, this second impact will be felt.
Now I believe very strongly that the first of these
applications is by far the most important: the influence
of computers on the whole urban environment. The
technologist might disagree and consider the other application far more interesting. Thoughtful consideration; however, may cause him to alter his views. Next
to the issues of war and peace, themselves, the problem
of whether we can create a livable urban environment
is certainly very high on the agenda.
There is also reason to be very pessimistic. Looking
at our traffic-choked, polluted cities, for all of our
advanced technology, for all of our wealth, we live in
kind of a squalor. Frank Lloyd Wright once said that
America seems to have an unrestrained, impassioned
Just for ugliness. I don't like to be unkind to our host

The impact of computers on urban transportation

city, but if you walk around Las Vegas at nighteven a person who has very little aesthetic sense can't
help but question what our wealth and technology have
done to our urban environment.
The reasons behind this decayed environment will
make it obvious that the computer is not just a vehicle
for the automation of existing operations. Its role
rather is so fundamental because our urban environment has gotten out of hand.
The causes for this unfortunate situation are two:
First of all, the urban environment, the metropolis,
has become so complex that it's impossible for any
person, architect, plariner, economist, or anyone else
to really grasp and to create a design to fill the objectives that everybody seems to desire. Secondly, even
if he were able to design such a city as the grand
master planners of old envisioned, even if he were
able to comprehend this complexity, he could not do
anything about it because urban decision making and
control have also deteriorated to the point where no
individual or group can really influence the shape of
the changing environment. It isn't just a question of
political power. Even if this power was given to those
people urging metropolitan government, it really
couldn't be used intelligently because of our sparse
knowledge of the urban growth process. The computer
has arrived none too early. Let us examine its current
role in urban design.
Transportation and the urban land pattern are intimately related. At the subdivision level or in a great
metropolis, common sense logic supports the idea that
the location of activities and the circulation between
these activities is interrelated so that the problem of
urban transportation planning is really a problem of
total urban design. Attempts to solve this design problem have gathered together an interesting group of
people in metropolitan transportation studies. City
planners, traffic engineers, highway engineers, systems
and computer specalists have formed a multidisciplinary team. Some progress has been made. To understand the reasons for this progress, a discussion of the
methodology of these studies is appropriate.
The first task in the urban transportation planning
sequence is one of data collection, processing, and
analysis. A large quantity of information is a prerequisite for intelligent planning. The land use pattern
must be determined in great detail. The detailed land
use pattern has never been measured in most urban
areas. So it is usually necessary to start a new data
acquisition program. Data on the travel habits of
people are also needed. This data is obtainable only by
actually interviewing individual people to find out the
origins and destinations of their trips in order to es-

tablish a pattern for projecting travel patterns into the
future. Other data on resources such as soil and water
must also be collected because of their limiting effect
on the final plan.
An idea of the scope of this data collection may be
inferred from experience in Southeastern Wisconsin,
an area with about 1.6 million people, which is actually
one of the smaller metropolitan areas. The basis planning information in this area involved 200 to 300
million characters of data. The area served by the TriState Planning Committee in New York City requires
a data bank of 2 to 4 billion characters of data.
That computers have made it possible to even
consider handling this vast quantity of data at all is a
t~uism. But serious problems in the form of data
processing "bottlenecks" have appeared. Typically
these studies start with high hopes. A large scale data
collection is the first step. Some areas never have been
able to successfully digest the data collected and the
planning process slows to a halt. Although the computers have raised bright hopes and fostered high
ambitions, the weakness of current information retrieval
and analysis software has tended to dim these hopes
and discourage excessive ambition.
Data handling in these urban transportation studies
has the disadvantages of both business data proccssing and scientific computations. Extensive computations are accompanied by a large quantity of input. The
lack of sophisticated information retrieval and analysis
techniques has led to the use of crude "brute force"
processing practices.
The second task in the planning sequence is that of
forecasting. In order to make a plan for an urban area
it is necessary to look at least 20 to 30 years into the
future. The problems of future land commitments and
the life of transportation facilities extend the time
horizon of urban plans. Forecasts of land requirements
reflecting quantity and quality are also needed to provide for future transportation facilities.
In forecasting, the role of the computer has been in
support of economic forecasting models. In fact, next
month I will be a member of a panel in Washington,
D.C., where the question will be asked: "Have mathematical models and the computer solved our forecasting problems?" The answer will very definitely be
no, but a start has been made. These models, although
far from perfect, have forced a rethinking of the
economic variables and relationships related to forecasting.
The next phase of the planning process is that of
design. Very little has been accomplished to date in
this area, but it is here that the most important role
for the computer probably lies. After all of the in-

The impact of computers on urban transportation
formation has been collected and processed, after the
forecasts have been estimated, the central problem of
transportation planning or, more inclusively, urban
design still remains.
The urban design problem may be defined as follows: Given certain objectives (the varied needs of
the urban population) , given certain constraints
(technological constraints and human constraints) and
given the related costs, how do we design an urban
pattern and related urban transportation and utility
facilities? Despite the large amount of data collected
by urban planning agencies, this data is rarely used in
a direct manner in urban design. It is not possible
with existing analytical techniques to use the data
directly in plan synthesis. The relationship between
urban design and systems analysis has been very
indirect. It would seem that all of this data collection
analysis and forecasting will do little good unless it
can be integrated into plan design. I, and a number
of other urban system analysts in the United States and
Europe, are now involved in an effort to develop
mathematical programming type models in which it
will be possible to quantify objectives, estimate costs,
insert constraints and provide urban design patterns
to aid the planner in the critical function of urban
design.
The last function in planning where computers have
had effect and have made great progress is that of plan
simulation or test. Simulation models of highway and
transit networks have reached a high state of development. Models exist for forecasting the number of trips
that will be generated by residential, industrial, or commercial areas. Other models will then distribute this
traffic by origin and designation using certain gravitytype formulas. Still other models will assign this traffic,
both with or without a capacity constraint, to certain
freeways or arterials in an intuitively conceived system.
These kinds of traffic models have been applied in
urban areas. I think this is greatly to the credit of
certain federal agencies, mainly the Bureau of Public
Roads, that have developed very comprehensive program packages for the application of these models. Despite these noteworthy advances, these simulation models can only test intuitively conceived networks and do
not aid directly in the design process itself. Future development of design models will probably eventually
obsolete this class of simulation test models.
So much for the application of computers to system
planning and design. The other function of the computer, and up to this time not a very important one,
but one that has great promise, is the use of the computer as part of the transportation system itself. In
both types of urban transportation, transit systems and

individual vehicle systems, the role of the computer is
becoming more apparent. Computers are being tested in
the roles of vehicular controllers as well as traffic controllers. The roles in planning and design are not independent areas. An example of their interaction in
transit is illustrative.
There's a very strong political and even emotional
pressure in the United States to persuade cities to adapt
rail transit systems. That a traffic problem exists in most
large cities is obvious. There's a great tendency, however, for people, not only politicians and the general
citizens, but even people with technical qualifications
to ignore the nature of the overall problem of the city
and its needs and to question whether these transit
systems are even economically feasible much less socially acceptable or aesthetically desirable. Some studies
in Southeastern Wisconsin have indicated that the free
market has already provided an economically sound
solution for urban transportation. Fancy rail transit
systems, automated or not, often encounter market
difficulties because they do not fill any previously neglected market need. For a new transit market to emerge,
the pattern of the city or the metropolitan area itself
much change. So there's a very strong interaction between the basic design of the urban pattern and the
role of the computer in transportation systems. Operating applications of computers in the system may be
discussed in two time-sequenced phases.
Looking at the here and now, the first applications
in transit systems are under way in the Bay Area Rapid
Transit System (BARTS) near San Francisco. Four
separate companies are testing control systems, all of
which involve computers to a greater or lesser degree.
The most complete system includes all three separate
uses: in vehicular control, in transit scheduling from a
series of satellite computers, and finally in overall control from a centralized computer. The second application, in dispatching, schedules the transit system in such
a way as to provide service with minimum cost, and
the third is in the more conventional areas of business
data processing extended most noticeably into fare
collection. These systems are now being evaluated.
Only one of the four systems actually involves using a
digital computer for all three of these roles. Two use
analog computers for vehicular control. Two of them
use an additional computer for dispatching, and one
of the systems applies analog computers to vehicular
control dispatching, and a digital computer only for
business data processing. BARTS is very definitely a
feasible application. There's no question about it, and
the real question here is only which is the best system
from an operational as well as a maintenance point of
view. It is easy to exaggerate the importance of the

The impact of computers on urban transportation

economic effects of transit automation. Automation
does play a role in the transit economics, but it is becoming pretty well understood that this role is not
economically decisive. In other words, if a profitable
transit market does not exist in an urban area, automation is not going to change the situation. There was
a great excitement a few years ago over monorail systems. There are various other emotional flurries at
times towards other transit system proposals, but even
with the aid of the computers and automation, the
limiting factor is still an adequate market demand in
most urban areas. A minimal level of transit demand
is required to support a transit system. The computer
is not likely to influence this basic market demand.
Another important "here and now" application of
computers is in traffic control systems. Some years ago,
the first attempt at traffic control using computers was
initiated not in the United States but in Canada. Traffic
Research Corporation developed a system in Toronto.
This system has been extended to the control of 300
intersections. These traffic control systems, by controlling traffic signal cycling in response to volume and
spacing of traffic, can actually increase the capacity of
systems as much as 50 or 60 percent. High hopes exist
for a system being developed for New York City. The
basic limitation in such systems does not lie with the
sensing equipment, the communication equipment, or
the computer, but, as in most process control applications, with the knowledge of the process characteristics.
Little is known about traffic flow as a process. Although
it is possible to simulate transportation systems in an
aggregate sense, knowledge of microscopic traffic flow
is still very slight. The success of the traffic control
system in Toronto has been· dependent, I think, on
experimental knowledge not based on a theoretical
model of the traffic flow process.
Looking into the future in the light of present accomplishments, it is evident that applications to date,
except in urban design, will probably not revolutionize
urban transportation because they really do not relate
directly to the basic problem of urban transportation.
,There is much 'concern about this urban transportation
problem today. The HHFA (Housing and Home Finance Agency) has been raised to a cabinet-level department. The HHFA has sponsored a number of
transit demonstration grants. The success of these programs has been somewhat questioned. President Johnson has recently placed great emphasis on transportation with his recommendation of a cabinet post for this
function. New studies are aimed at the formulation of
a National Transportation Policy. A major program
has been initiated for the northeast section of the U.S.
known as Megapolis. The Department of Commerce

has recently sponsored a series of studies at Cornell
and MIT. The Bureau of Public Roads is also developing studies in this same area. Finally, all of these
studies must solve the one basic problem. Amidst all
of the hand-wringing and arm-waving, it really reduces
down to one fundamental problem: urban areas. The
way such areas are developing is fundamentally toward
very low-density, sprawling-type development. With
this type of development, the automobile, however horrible it may seem to some planners and other people,
is actually an ideal form of individualized transportation. It may not seem desirable at times for the community, but it is for the individual. People can complain about congestion, air pollution and other ills,
but until some alternative to the automobile is suggested (and it certainly will not be in the area of rail
transit), until some system is developed which has
flexibility, which faces the problem of air pollution,
and which is able to provide something equivalent to
the automobile, the situation is probably not going to
change much.
Urban transportation studies have now analyzed a
large amount of data to discern the nature of transitriding. They have discovered that many of the people
who utilize transit have no other choice. They are
known in the trade as "captive riders." Given a free
choice, many of these riders would not choose transit
most of the time. Some of the new transit systems being suggested do not seem to face the captive nature of
transit riders. Future increases in transit ridership must
come from the noncaptive, who must be sold on the
benefits of transit travel.
Some of the new designs are not only technically
weak but psychologically unattractive. Vehicles, that for
all of their automation technology still resemble a New
York Subway Car (with standees during the rush hour)
are probably not going to convince many people that
they should abandon their automobiles for the transit
car.
If there is a solution, what role will thc computer
play? Although no generally acceptable complete solution has yet been forthcoming, the general consensus
seems to involve an individual electric vehicle that can
operate in these low-density areas. The Cornell Study
suggested an Urbmobile. The Urbmobile could be driven
locally for pickup and distribution and also on automated freeways for travel over longer distances within
the urban area. Although this system is still at a very
conceptual stage, it at least has the potential for the
solution of the basic urban transportation problem that
involves high volume traffic in certain areas at certain
times, and simultaneously low-volume travel to a wide
dispersion of trip origins and destinations. A system

The impact of computers on urban transportation

such as the Urbmobile will obviously require computers for vehicular control and for traffic control.
In other areas of transportation, only partially urban
and mostly intercity, the computer will also play a
major role. The automated highway for intercity transportation will probably come along in the next 20 or
30 years. The problem is not as overwhelming as the
urban one and probably not as critical, but is much
easier to solve with existing technology. The problem of
terminal operations and freight transportation will also
very definitely involve the computer.
In summary, the role of computers in urban transportation systems will be important and sometimes
crucial. This role will definitely not be peripheral. People now seem to desire a better urban environment, but
the complexity of the urban patterns has need of the
computer to solve the problems of urban design. The
computer will also have a role in the operation of
transportation systems. To realize this potential of the

computer in urban development, two needs are apparent. There are some subsidiary technical needs in
data collection and data processing of computer software for information retrieval and analysis, but I think
these are very minor compared to the major problem,
the problem we have addressed ourselves to today,
namely, the marriage of analytic and substantive knowledge. The greatest technology can be very sterile unless
it is combined with substantive knowledge. I think the
experience of operations research and management
science in industry testifies to this need. Many operation's research projects have failed in application from
a lack of substantive reality. In urban problems the
need for this marriage of the professional, who is not
information technology-oriented, of the politician, and
of the information technologist is even more critical.
This marriage is the central problem that must be
solved if we are to realize the potential of the computer
into the field of urban planning and transportation
systems.

Panel Discussion Records

A Panel Discussion

The computer industry in the buyer's market

J. A. RICCA, Raytheon Computer, Chairman
J. P. ECKERT, UNIVAC/ Sperry Rand Corporation
W. J. GALLAGHER, Radio Corporation of America
R. W. HUBNER, International Business Machines Corporation
R. D. SCHMIDT, Control Data Corporation

SUMMARY

CHANGE OF TECHNOLOGY

The panel, in general, agreed that the computer industry is now in a buyer's market. There was one
notable dissenting point of view which indicated that
the technology is still rapidly changing, particularly in
peripherals, and thus, that the buyer's market would
not prevail until at least 1975. The following industry
characteristics appear to support the existence of a
buyer's market.

In the determination of whether the computer industry is still in a seller's market or has entered a
buyer's market, the state of the technology is pivotal.
To facilitate this appraisal, computer history has been
divided into three eras of technology as follows:
1. Early era-1943-1950. The industry was forced
to use existing components, such as vacuum tube
logic and Williams tube storage. There were considerable new system ideas but user requirements
were not very well known or understood.
2. Growing era-1950-1960. Considerable development of components specifically adapted to computers took place in this period, and the so-called
solid state computer emerged using transistors,
diodes and ferrite cores. Magnetic tape units,
disks and other mass storage units came into
being. Although the system concepts were not
basically changed, refinements were made in adding items like index registers and floating point
arithmetic.
3. Refining era-1960-1975. Discrete transistor and
diode circuits are now being replaced by integrated circuits. This trend has just started and
will continue for several years before IC technology stabilizes. There is considerable hope that
peripheral development will move forward at a
rapid pace and catch up with the highly developed
central processor. Refinements will continue in
system concepts but no fundamental change is
expected.

1. The rate of change of the technology has slowed
down considerably, and fundamental breakthroughs are becoming less frequent. There is
now a fair degree of engineering standardization.
2. Computer users strongly influence product design,
and the product planning procedure of designing
to meet user needs has become a basic marketing
tool in the industry. Thus, the user has a wide
variety of choices at attractive prices.
3. The nature of the competition is very vigorous,
and prices are getting lower. In addition, other
inducements to buy are common; such as software thrown in and offering of special services.
4. The number of competitors has been reduced in
recent years, and it is more difficult for newcomers to enter the race. The entry fee, already
high in the past, is getting even higher.
5. Massive investments are being made for tooling
and mass production of standard building-block
product lines which cover a large percentage of
user applications. The introduction of integrated
circuits is further accelerating this trend.

The state of the technology is changing and will
continue to change. The key issue is whether the rate
59

The Comuter Industry in the Buyer's Market

of change of the technology and the frequency of fundamental breakthroughs have created the characteristics
requested for a buyer's market. It is apparent that the
rate of change has slowed down considerably and fundamental breakthroughs are becoming less frequent. These
trends tend to create the climate required for a buyer's
market.
Another test is the amount of engineering standardization. When computers are being sold on a custom
engineered basis, the user pays for all the special engineering and software, and this situation favors the
seller. There is now a fair degree of engineering standardization in the industry, and the manufacturers have
heavily committed engineering and software expenditures in advance of offering the product line to the
marketplace. These practices favor the buyer.
USER INFLUENCE IN PRODUCT PLANNING
Now that the user requirements can be defined, the
traditional product planning process has taken over.
Each manufacturer solicits user groups and structures
new research and development projects that will result
in products that satisfy these user needs.
Because of the relatively large number of manufacturers, the user has a wide variety of choices at attractive prices. This is an essential ingredient for a
buyer's market.
VIGOR AND COMPETITION
The nature of the competition is very vigorous, and
this supports the existence of a buyer's market. It can
be argued that the number of competitors has been reduced in recent years and it is more difficult for newcomers to enter the race. The entry fee, already high
in the past, has gone even higher. Nevertheless, the quality and the strength of the remaining companies has
increased. There are now several competitors who can
compete with IBM on a broad scale basis.
Because of the highly competitive situation, prices
and profit margins are lower. This would appear beneficial to the user, at least on a short range basis, but
could be harmful on a longer range basis if research
and development spending for new products were curtailed. This does not appear to be the case, for all companies are maintaining or increasing their research and
development expenditures. In order to get back lost
profit margin, a general attack is being made on manufacturing costs. Also, continued industry growth will
spread the research and development investment over
more dollars of income.
In addition to outright price cutting, other related
sales inducements have become common. A particularly

popular one is "throwing in" software. Special services
are also provided as the occasion may warrant.
TREND TO MASS PRODUCTION
As the rate of change of the technology diminishes
and engineering standardization increases, a new form
of competition is developing in the manufacturing area.
The new "building block" product lines and the use of
integrated circuits encourage huge investments in plant
and tooling aimed at lowering manufacturing costs and
improving the reduced profit margins. Because of these
huge investments, the life cycle of new models will
lengthen, and engineering changes will occur less frequently. The trend towards mass production is a clear
indication of the maturing of the buyer's market.
MANAGEMENT OF INDUSTRY:
SCIENTIST VS. BUSINESS MANAGER
In the early period of the industry, the scientist
emerged as the dominant managerial figure. The key
problems were mastering the new technology and building computers that would work with reasonable reliability. User needs and requirements were of secondary
importance. Now that the pioneering days are over and
computers have reached a high degree of reliability, the
business aspects of the industry have taken on increased significance. The business problem is generically difficult because of the complex nature of the computer system itself, the impact of leasing and its requirement for keeping the installation sold over a period
of time, the difficult installation and servicing problems
of computers and the vast size and growth of the industry. To cope with this problem, industry must develop top managers who are both technically and business oriented. The engineer turned generalist must acquire business knowledge and acumen in order to manage. Similarly, the businessman who comes up from
marketing, finance, or other business disciplines must
acquire enough technical orientation to successfully
manage the business. The development of this new
breed of manager has become one of the critical problems of the industry.
The Technical Program Committee invited other
members of industry who were not represented on this
panel to add their views. The views of Mr. Max Palevsky, President, Scientific Data Systems, Santa Monica,
California, are appended below:
A discussion of the computer market place that is
not cast in terms of the influence of IBM in structuring
that market place has the ring of the discussions about
the Emperor's new clothes. The premise of the discussion at the 1965 FJCC-that a basic change has oc-

The Computer Industry in the Buyer's Market
curred in the computer market-seems to have little
relationship to the relevant facts. What are the relevant
facts? Every industry has cyclical fa~tors which affect
profit performance. These include demand, plant capac:..
ities, technological innovations, etc. When demand is
high relative to supply, the market is said to be a seller's
market. The only measure of this relationship is general profitability. The computer industry, since its inception, has never had a period of general profitability.
The total losses, excluding IBM, are astronomical and
continue to grow each year in spite of constant mutterings about "turning the corner." Probably no other industry of the size and importance of the computer industry has experienced an era of rapid technological
change in which a seller's market never existed. Recent
examples of industries that have experienced rapid
change and have had a consequent period of general
profitability are color television and semi-conductors.
If such a period will ever exist in the computer industry
appears open to question. The cash flow of IBM is of

the order of $1 billion a year which dwarfs the strength
of even IBM's largest competitors. The discussion of
other factors affecting the market place-such as the
rate of technological change-are not totally irrelevant
but are of secondary or tertiary importance. Because
of the character of the corporations that are- IBM's
competitors, there has been a reluctance to openly discuss the problem of economic concentration. For a
professional organization, however, to avoid the central
issue is not in the best interests of the industry in the
long run. Obviously there is a limit even in American
industry to the investment that large corporations are
willing to make unless there is a strong probability of
return. Without such investment on a broad scale with
many competing firms, the potentialities of the computer field will never be fully realized. That is the central issue relating to the kind of computer market we
have and the consequences that can be predicted. If
this central issue is not treated, the discussion is really
about the Emporer's new clothes.

A Panel Discussion

The overseas computer market
Chairman:
MILTON C. MAPES, JR., Deputy National Export Expansion Coordinator
U.S. Department of Commerce
Panelists:
DONALD F. ORR, Vice President of International Operations
UNIVAC Division of Sperry Rand Corporation
JAMES G. MILES, Vice President for Marketing Development
Control Data Corporation
NORMAN J. REAM, Director, Center for Computer Sciences and Technology
Bureau of Standards, U.S. Department of Commerce
THEODORE L. THAU, Executive Secretary, Advisory Committee on Export Policy
U.S. Department of Commerce

MR. MILTON C. MAPES, JR.

chines. On this basis the United States has 39% of the
total world market in international commerce, followed
by West Germany with 14%, the United Kingdom and
Italy each with 11 %, and France and Sweden with
about 9% each. That includes accounting machines,
which still comprise a very large part of the market. I
suspect that if it were limited to computers alone the
U.S. share of the world market would be. substantially
greater. As to where our computer exports go, I also
have some pertinent figures. In 1964 our major cus-

Our procedure will be to devote the first half of the
discussion to presentations by the panelists on various
aspects of the overseas marketing problem, and the
second half to questions and participation from the
floor.
I know many of you are deeply involved in export
marketing and the international business aspects of
computer sales; in 1965 U.S. computer exports are
going to run approximately 400 million dollars. Our
imports at the same time are running approximately
60 million dollars. It might be significant to mention
that computer exports are up almost 300% from the
1958 figures, only seven years ago. At that time our
computer exports totaled only $103 million, as shown
in Table 1.
TABLE 1. Computers-Exports and Imports
(values in millions of dollars)
Exports

Increase in
Exports

TABLE 2. U.S. Exports of Computing and
Related Machines

(values in millions of dollars)
Percent
Change

1 963
21%

Japan
$63
United
Kingdom 35
Canada
39
France
41
West
Germany 38
5
Australia
Ita]y
8
Switzerland
8
Sweden
6
Other
countries 59

Imports

(percent)

1958
1963
1964
1965 (est.)
1966 (est.)

$103
302
369
400
445

193
22.2
8.4
11.3

$15
57
57
60
65

The United States' share of the world computer market is a little difficult to determine. I do have a figure
which includes both computing and accounting ma-

$302

$71

1963-64
130/0

62
55
38

17
15
10

41
-7

2
3

28
20
11

8
5
3

-26
300
38

3
2

9
7

19
100%

12
13

17
100%

$369

SOURCE: Bureau of the Census.

1 964
19%

The Overseas Computer Market

tomer was Japan, which took 19 % of our exports; our
second major customer was the United Kingdom with
17 %, and the third was Canada with 15 %. The worldwide distribution of these exports for 1963 and 1964
is shown in Table 2.
I'd like to emphasize that the potential for growth
of this market in the future is tremendous. For example,
the total market for computers in Europe in 1965 is
estimated to be about $450 million; by 1970 it is expected to be in excess of $1.5 billion. Our problem
is how to penetrate that market most effectively. The
United States, if it intends to maintain anything like
its present share of the export market, is going to have
to get out into the world market more than ever before.
The tendency in the U.S. is to concentrate on developing
the domestic market. There is very little tendency by
most businessmen to export. When you talk to businessmen who have 12 to 18 months backlog tied up with
orders merely from the United States, you can't interest
them in getting into the intricacies and what may appear to them to be strange procedures of export marketing. I think though, that in the gradually growing
one-market world which we find coming upon us with
increasing speed, the United States businessman is
going to have to get out and sell on the main streets of
the rest of the world if he is going to compete on the
main streets of the United States itself.
I want to discuss briefly the Export Expansion Program in general terms, and then Mr. Orr will present
an introduction to the export of computers. We have
had serious balance of payments problems, as you have
probably read, consisting of a deficit in our balance of
payments. Every year since 1950, except one, we've had
a deficit in our international financial account. Since
1958 our gold reserve has dropped off from more than
$22 billion to less than $14 billion at the present time.
This is particularly significant because the dollar is the
primary monetary unit in international exchange, and
since that is backed up by the Fort Knox gold hoard,
if that becomes substantially lower, or if there should
be a lack of confidence in the dollar and unWillingness
by foreign nations to hold the dollar as backing for
their currency as we hold gold, it could result in an
international monetary crisis and a serious collapse of
the entire international financial structure. So we have
inaugurated a voluntary program limiting overseas investments by U.S. corporations, which has been a subject of considerable discussion. We have also had a
voluntary program administered by the Federal Reserve Board to limit all foreign lending this year to
105 % of the total of all loans outstanding at the end
of last year.

Then we have the National Export Expansion Program, in all its aspects. First, we have the services of
the Bureau of International Commerce in the Commerce Department. I am running through these because I think many of you may not know what services are available to you in the international market
directly from the Commerce Department. The Bureau
of International Commerce has programs carried out
by its offices of International Trade Promotion, International Region Economics, International Investment, and Commercial and Financial Policy. We have
permanent trade centers in six major cities around the
world. These are permanent exhibits of U.S. products
and are always available for foreign buyers and businessmen who want to learn what is available from the
U.S. Those trade centers are in London, Tokyo, Milan,
Frankfurt, Bangkok, and, for the first time this year, in
Stockholm. We also have business information centers
all over the world and U.S. foreign service commercial
officers. There are presently 156 commercial officers
with the U.S. Foreign Service, and their work is coordinated by the Commerce Department. Their function is to service American business overseas. We have
increased the budget during the last three or four years
for trade fairs and trade shows and the Commerce
Department has promoted U.S. commercial exhibits.
We have trade missions, consisting of United States
businessmen, which go over with Commerce assistance.
They represent the entire United States business community for their own specialty and arrange for agency
relationships, overseas branches, licensing agreements
and even initiate negotiations for joint ventures. These
businessmen do not represent only themselves. They
represent a broad sweep of the firms involved in their
line of business. These trade missions have been extremely successful in establishing relationships and
doing business for U.S. {frms overseas.
We have in the United States 42 field offices of the
Department of Commerce. I would urge any of you
who are interested in getting into the international
trading community and have not done so, that your
first point of contact should be the field· office nearest
your home base. It can supply you with information,
printed material and very genuine technical know-how
on how to trade overseas.
We also have 42 Regional Export Expansion Councils, which are composed of local businessmen in each
of the regions served by the Field Office. They have
an operation whose purpose is to increase the participation of businessmen in foreign trade. The name of it is
Operation 10,000-the object being to get 10,000 additional businesses into the overseas market. In addition

The Overseas Computer Market
we have the National Export Expansion Council which
is composed of top level businessmen from all over the
country. It recently appointed three action groups. In
the last two months these groups have been studying
specific problems-one doing export financing, one
studying ocean freight rates, and the third working on
tax incentives fQr export.
The Export Expansion Program is largely not subject to being programmed in the electronic sense. One
major exception to this was recently acknowledged,
when the Business and Defense Services Administration
in the Commerce Department established a computer
program to bring together overseas trade opportunities
and the export capacity of United States firms. Approximately 50,000 U.S. manufacturers are now registered for this program. The idea is that as trade opportunities come in from overseas they will be mailed to
those equipped to handle international trade. Perhaps
much more will be and can be done in the line of automating and programming the export effort. At this time
not very much has been done.
I ran across an interesting quotation the other day
by Dr. Herbert Simon in his book on automation discussing the problem of the general problem solver. He
stated: "Problem solving proceeds by erecting goals,
detecting differences between present situation and
goal, finding in memory or by search tools or processes
that are relevant to reducing differences of these particular kinds, and applying these tools or processes.
Each problem generates sub-problems until we find a
sub-problem we can solve-for which we already have
a program stored in memory. We proceed until, by
successive solution of such sub-problems, we eventually
achieve our overall goal-or give up."
Our overall goal here is to increase our overseas
sales of computer materials. Fortunately we can break
this up into sub-problems by areas and types. This panel
has been designed primarily to do this.
With that introduction, I want to introduce Mr. Donald F.
Orr, Vice President for International Operations of Univac
Division of Sperry Rand Corporation. Mr. Orr is in charge of
all the Division's international operations (engineering, marketing, and manufacturing) overseas. He has been in New York
for the past five years. His previous 13 years were spent overseas with Sperry Rand and its predecessors. He is a graduate
of the George Washington University School of Foreign Service. His subject is going to be "A General Introduction to Computer Marketing Overseas" with emphasis on Europe, both
East and West, and the developing countries. Mr. Orr will also
discuss some specific problems relating to the import of computers.

MR. DONALD F. ORR
During the past 13 years, the number of computers
in the world has grown from a few to over 35,000 sys-

tems. In Japan and Europe alone some 7,800 computers
are in use today, representing a value of $1.3 billion.
This market is growing at the rate of about 20 % a
year, and it is expected that more than 22,500 computers will be in use in this area by 1971, representing
a purchase value of over $4 billion.
The steady growth of the European market should
bring it to about the same level as the United States
within another decade. This is on top of the fact that
because of the traditionally lower labor costs in these
markets, the economic savings that a computer installation offers a businessman are not quite as easily justified as in the United States. Therefore, as a general
rule I have found among European and Japanese users
a relatively high degree of sophistication in the use and
application of their computers.
If anything inhibits the growth of the computer
market in Europe, it will be the shortage of qualified
programmers and operators. At this time there is a
need for 120,000 people with computer knowledge,
and by 1971 this demand is expected to rise to 300,000.
The U.S. computer manufacturer will face increasing
competition from the local manufacturers in a growing
number of the foreign countries where we are now selling our products. These competitors have indeed mastered the art of building computers and in some cases
their technological designs are in advance of our own.
To be competitive we must be fully knowledgeable of
the hardware and software needs of these markets and
understand the specific requirements of our overseas
customers. We should be prepared to build these needs
into the products to be shipped to these areas.
While export shipments of computers from the
United States to other countries are higher than ever
before, the ability to import computer systems into
many of these countries at the same time becomes more
challenging as time goes by. The changing tariff picture in Europe, the emergence of the European Economic Community, restrictions on importations of certain sizes of computers into Japan, are but a few of the
factors which oblige the U.S. manufacturer to seek new
means of maintaining his share in these markets and
to be able to continue to support present customers.
Such steps have brought about a variety of arrangements such as joint ventures and licensing for manufacturers, prevalent in Japan, mergers and partnerships
with foreign computer manufacturers, and the establishment and expansion of wholly owned U.S. plants
overseas. In many cases such actions appear contrary
to the government's program to improve our own balance of trade as well as correct the present balance of
payments deficit through voluntary restraints on new

The Overseas Computer Market

investments abroad. Noone country, or no one company, for that matter, can feel secure that it has a
permanent lead in this fast-growing computer industry
that is continually being pushed forward by major
technological breakthroughs and the demands for its
products. Therefore, the U.S. computer manufacturer
is obliged to take such action as is necessary to maintain his present position in these markets.
To meet the challenges the overseas markets offer,
our industry is continually seeking new ways of financing our export business. In this regard, for example,
some provision for financing the leasing overseas of
U.S.-manufactured computers would be helpful.
Turning to Eastern Europe for a moment, the lure of
large potential markets within the Soviet bloc has been
getting increased attention from western manufacturers.
While it can be assured the need for computers in these
countries may be as great as in Western Europe, in
proportion a relatively small number of computers are
in use behind the Iron Curtain today. Most of these
have either been supplied from Russia or from firms in
Western Europe.
Up until now, our own Export Control Act, which
classifies computers as strategic goods, has restricted
U.S. firms from doing business in Eastern Europe while
Western European manufacturers freely export equipment of comparable technological design into these same
countries. Perhaps we may hear a little more on this
subject from my co-panel member, Ted Thau, later
this morning.
Building a computer market for the less developed
areas in the world should present us the greatest challenge of all. This area encompasses parts of Central and
South America, the Middle East and Africa, and the
Far East outside of Japan and Australia, excluding, of
course, the Chinese mainland. These markets are indeed far flung on the map. They represent more than
40 % of the total world land area with a population of
1.2 billion people. This area has probably the greatest
need for computers, but at the same time is the least
prepared to use them effectively.
Mr. U. Thant, Secretary-General of the United Nations, has said, and I quote: "The computer is the
means by which the developing countries can bootstrap
themselves to reach the technological level of the industrialized countries." Almost without exception, each
of these areas is caught up in a fast-growing internal
economy coupled with the entry into world trade on a
competitive basis that is calling for the need for better
controls, lower costs, and more efficient ways of conducting business. Electronic information processing
techniques, through the use of computers, will play an

important role to bridge the gap between their centuryold business practices and the modern methods of
business administration, control and decision making.
Before this can become a reality however, there must
be faced up to and resolved the problem of a severe
scarcity, and in some areas an almost complete lack, of
qualified people who can be trained to make use of
computers in business. In commenting on this, E. Dinah
Gibson of the San Diego State College, who has studied
the problem, stated that many of these countries are still
not even teaching business administration in their universities or other institutions of higher education. This
is a "must," he feels, as this is the base on which
business data processing must be built. The development of the computer market, therefore, will be limited
to quite an extent by the business education background of business executives at all levels until a way
can be found to provide this background and ability
for them.
The manufacturers are still the original trainers in
many of these less developed areas. Several companies
have set up training facilities such as in North Africa
where customers and prospect personnel may come to
be indoctrinated. This is not enough as it does not go
far enough. It is my suggestion, therefore, that through
our own government and/or U.N. sponsored programs,
together with educational bodies of the more developed
nations, we join together with the governments and
centers of learning in these lesser developed countries
to provide a solution to this educational program for
training people not only in electronic information processing techniques, but in the basic concepts of business
itself. This program would be a big step forward in
enabling these countries to make the technological and
economic progress that is essential to their citizens'
well-being. I am sure that the various computer manufacturers throughout the free world could make a worthwhile contribution in one form or the other through
such a program and would be ready to collaborate if
called upon.
At the present time it is estimated that there are less
than 350 computers in use in the less developed areas
about which we have been speaking. These are mostly
being used by government agencies and in some universities and by the larger, internationally oriented,
foreign-owned companies. This total could well increase by tenfold in the next 10 years, provided skilled
personnel are available.
The establishment of adequate sales, support, and
service facilities on the part of the manufacturer in the
less developed countries is vital if he expects to successfully compete in these markets. This will call for in-

The Overseas Computer Market
vestments of both money and talent. The volume that
could be anticipated from such a market and whether
this can be profitably obtained will be an important
factor in making any decisions to set up these facilities
in the first place. Since we are dealing in areas which
may be plagued with unstable currencies, higher rates
of borrowing and other requirements such as prior
deposits, restrictions on remittances and so forth, a
means of sound financing of our· computer export sales
to these countries is a very important factor.
As a means of sharing part of the risks of doing
business in these areas, local distributors may be the
answer. They would be fully acquainted with the local
market, understand their laws and business customs,
and, hopefully, provide the necessary local investment
and the financing of resulting sales. Joint ventures,
made up of the manufacturers and the local interest,
thus assuring the manufacturer of a share in the development of the business as well as overseeing sales
and service standards, is still another approach. At
Univac we have been successfully operating overseas
using a combination of these methods in addition to a
substantial number of wholly owned subsidiaries and
branches.
Day to day operations are another thing. Underdeveloped and overloaded communications will restrict
the applications of real-time techniques and other features that American manufacturers can offer in their
computers today. Software, as we provide it, will not
necessarily meet the requirements of these areas and,
if anything, should be made more easy to use than it is
today. Electrical power requirements are different and
vary even between locations in the same country. Such
concepts as operations research, CPM, and other management aids are not being applied and even less widely known. We can expect to have the computer operating in environments far below what we would consider
ideal, or possibly even acceptable, in the United States.
Unstable power supplies, problems of heat, dust, and
humidity, are a few of the situations that the computer
exporter may be faced with, depending on the country.
Spare parts backup and logistics involved in supplying and maintaining these inventories in the respective
countries where computers are installed is an effort
which gets special attention at Univac. Many of these
parts are carried on the strategic commodity classification list and require export licensing to ship from the
United States. So that we may better serve our overseas
customers we are hopeful that the Department of Commerce will see fit to liberalize some of these regulations,
particularly for shipments within the western world.
The challenges and opportunities that lie ahead for

the American businessman in the less developed countries can be shared with others besides the computer
manufacturers. I envision an important role that can
be played by professional groups. such as the independent consulting firms and the EDP service organizations. There is a real need for "turnkey" type of services
to be made available and offered in these areas, which
would provide a potential user with a complete EDP
systems service encompassing all aspects of problem
definition, development of procedures, system installation and operation of the system until local capability
has been trained to take over. With the resources and
experience behind these professional groups, this should
be a natural as well as profitable expansion of their
business.
I also envision and encourage the leaders of the
emerging nations that already possess EDP know-how
to pool their financial and human resources and establish national government computing centers in their
respective countries, thus providing electronic information processing techniques on a "utility" basis to all
segments of local government and commerce. In this
way, all may reap the benefits that computers offer.
There is no doubt in my mind that the overseas market offers a real potential for American computer products in the course of the next few years. Our courage to
face up to the challenges these markets present to us,
and our ability to meet the competition from foreign
computer firms can represent an important contribution to the President's program for export expansion,
which Mr. Mapes has mentioned earlier, as well as
provide new profits for the American exporter.
MR. MAPES
One of the founders of Control Data Corporation eight years
ago, James G. Miles has a Bachelor of Science in Electrical
Engineering from the University of Nebraska. He worked on
radar development during World War II and somewhere along
the line picked up an LLB from St. Paul College of Law.
Jim's first subject today is going to be marketing to the countries of Eastern Europe. As Don Orr mentioned, there are a
number of governmental problems involved here. We will
hear more about them later from the government's side.
Secondly, Mr. Miles will discuss problems of marketing in the
developing countries, and third, the potential for computer
sales as devices for impact on the less developed countries,
promoting their development and promoting the interest of the
free world in these areas.

MR. JAMES G. MILES
The people sitting in this room this morning are
privileged to be associated with probably the most dynamic business in the world today-the making, selling,
and using of computers.

The Overseas Computer Market

Probably never before in history has a single set of
tools been developed which inherently contains so much
potential for constructive use for all people in the world.
We are making extremely good progress in the use of
these tools in this country, a fact which is well publicized. And the rest of the world observes, and all
desire the same for their countries.
You are all aware of the literally thousands of facets
of our economy and lives in which computers are
playing major roles today-from education to forest
management, from heart analysis to steel mill controls,
from banking to advanced communications. systems,
and on to space systems. And soon, the use of computers in total management systems to permit the optimization of our business enterprises will be achieved.
And the rest of the world observes and desires the
same for their countries.
There are several ways to look at this: Consider such
a large capability and such a large technological lead
and such a monopoly in the computer power of the
world (in this regard I refer to the fact that 95% of the
world computer market is vested in U.S. manufacturers)
as carrying with it a proportionate responsibility. We
are, in effect, the inheritors of a wonderful set of talents
in this country-a combination of creative energy,
financial capability, and a free enterprise system which
has permitted this. By virtue of this extreme good fortune, we are, or should be, in the position of beneficial
trustees. As a corollary, with this goes an extraordinary
set of responsibilities which virtually place, or should
place, their owners in the position of "trustees" in the
legal philosophy. These talents, under this· context, are
not entirely the property of the presumed owners to
hoard and dispense at their sole pleasure without consideration for others.
When we discuss the export of computers, and the
export of computer technology, we are discussing a
combined set of problems which probably offer more
opportunities, and which are at the same time complicated by more problems than almost any other export
subject that could be discussed today. These problems
are not all internal to the United States.
• They cut across the technological capabilities of
many countries to make and use computers, as
Mr. Orr indicated.
• They involve complicated problems of international finance.
• They involve the need (in the less developed countries the desperate need) for computers to help
them organize, build, and operate their economies.
• They involve the laws of the United States, and the

•

laws, policies, and objectives of many countries
who want to import our computers, technologies,
and our data processing techniques.
These problems heavily involve the foreign policy
objectives of the United States, and heavily interact, on the other hand, with the aspirations and
foreign policy objectives of many other countries.
And, when the U.S. attempts to put undue restrictions on the exports of these commodities, we
often obtain very serious reactions from the other
countries.
These problems involve also multinational trade
agreements such as the COCOM (Coordinating
Committee) Agreement, which in many respects
are more realistic and liberal than the laws and
policies of the United States.
And these problems involve a "balance of power"
in technology which can simultaneously work both
for and against the United States, depending on
how we handle it.
Above all, these exports, or their denial, offer the
greatest of opportunities for orienting many countries toward the United States, or otherwise.

Mr. Orr discussed the need for computers in less
developed countries and the widening technological gap
which is occurring between us and the less developed
countries. This is a very real thing. These countries
aren't even keeping up, in most cases, with their growing populations and in their ability to feed them, house
them, and provide them with the other amenities of
life which, by way of worldwide communication and
publications, they are easily able to discern is not their
lot to have, vis-a-vis the United States. We sit around
worrying about whether these same countries may become oriented with the Communist countries. We
wonder how and what to do about it. The consequences,
in my opinion, of this widening economic gap between
the less developed countries and the United States
should be of the utmost concern because the continuation of this problem, uncorrected, I believe, carries with it the seeds of increasing dissension and may
give these people reasons to believe they should orient
themselves with the Communist countries instead of
with the United States. So we really have a responsibility in this country to dig in and help these countries
to establish their ability to acquire a few computers and
get them trained along the lines Mr. Orr discussed. As
quickly as possible, also, we must help them develop
in-country capabilities to make these systems, so they
don't need to rely entirely upon us for exporting or for
importing.

The Overseas Computer Market

I am now going to discuss a subject which is highly
controversial-trade with the Communist and Communist-controlled countries. Various people of equally
good will hold deCidedly diverse opinions on this subject. Passions are keen. Some people think of the subject
in terms of black and white; they may be right. Personally, I and my company, Control Data Corporation, take a different view.
President Johnson said in his State of the Union
address in January 1965, and restated in a speech on
May 8, 1965, significantly marking the 20th anniversary of the end of World War II, the following:
Here is some of our unfinished and urgent business.
First we must hasten the slow erosion of the Iron Curtain.
By building bridges between the nations of Eastern Europe
and the West, we bring closer the day when Europe can be
reconstituted within its wide historic boundaries.
For our part, after taking counsel with our European allies,
I intend to recommend measures to the United States Congress
to increase the flow of peaceful trade between Eastern Europe
and the United States.

And again on May 23
Johnson said:

Lexington, Virginia, President

There is no longer a single Iron Curtain. There are many.
Each differs in strength and thickness, in the light that can pass
through it, and the hopes that can prosper behind it. . . .
We will continue to build bridges acros~ the gulf that has
divided us from Eastern Europe. They will be bridges of increased trade, of ideas, of visitors, and of humanitarian aid.

I think we should take the clue from this enlightened
attitude of President Johnson. Whenever I go abroad,
my main objective is to make friends for the United
States. I believe, among other things, that if I can
achieve this, the relatively simple process of taking
orders will essentially take care of itself.
Then comes the frustrating part.
I am told by people in Washington on one hand,
before I go on these expeditions, that I can pursue these
objectives without limits in all countries including Soviet
Russia, but excepting of course, Red China and Cuba
and one or two others. Then when we go out and
obtain some of these orders we come back and find
that there is no feasible w~ to obtain export licenses.
This is frustrating to the people in the Eastern countries
because, I guess, they presume by the very fact that I
showed up in their countries there is the opportunity to
do business. This does not too often turn out to be the
case when the export license is denied by the United
States.
Computers are, because of their inherent usefulness,
among the most valued; valuable, and useful items for

foreign trade. Consider the people they reach. First of
all, they reach the top people in each country. They
reach the top educators, the top scientists, and the top
businessmen; the top industrialists, and the top government people. There have been several countries in
which we have obtained large computer orders where
the head of state, the top man, has signed off on the
order before it has come to us. If you are building
bridges, or if it is your intent to build bridges, and it is
the intent to try to wean some of these countries away
from their past orientations (which, incidentally, many
incurred because of enslavement, and not because of
desire on their part), I believe that computers are the
most important commodities to be traded. I don't think
there is any particular point in going over and offering
to trade tables and chairs, because they can probably
manufacture such items as well or better than we can.
Besides, the trading of tables and chairs does not reach
influential people; such items are not matters of top
national concern. It's the people these reach that is the
important thing, and not just on Control Data's behalf,
either. As we would bring computer users into the fold
of our customers, they automatically join an international fraternity of computer users which exchange information and techniques. Fortunately most of this international fraternity is oriented toward the West politically and philosophically. These new users will become
associated with a myriad of users in the United States
and in West Germany, and in Scandinavia, England,
Australia, France, etc. I believe that if you want to build
bridges, this is the way. If you want to bore big holes
in the foundations of Communism, this is the best way
to do it, because of the influential nature of these tools,
and the people that they reach.
How do we proceed to obtain the conditions by
which this can happen, assuming that it is desirable.
Well, of course, as Mr. Thau is going to tell you in a
few minutes, there are many, many influences at work
in the U.S. in this regard. First of all, there are laws of
the United States which were established by the Congress; these laws include the Export Control Act, the
Battle Act, and several others; these laws establish the
basic framework of ground rules. On the other hand if
you read those Jaws, it is very interesting to see that
there appear to be many interpretations of those laws
which don't necessarily have to be read into them.
Oftentimes the Administrative Rules and Regulations
which derive under these laws are much more restrictive
than at least I read into the four corners of the words
of those statutes. I'm not the only one that holds this
opinion.
Then there are: international agreements, such as the

The Overseas Computer Market

COCOM Agreement which has rather set up some
terms and items which involve supposedly enforceable
multinational embargoes regarding certain so-called
"strategic materials" shipped to the Eastern countries.
Different countries put different interpretations on
these multinational agreements, but our country puts
the most strict interpretations on them, or at least imposes different sets of regulations which are more
restrictive-and, I believe, too severely restrictive.
Beyond that, of course, there are the executive functions of the U.S. government. The major departments
reporting to the President-Commerce, State, Defense,
Treasury and Justice-all get involved in these matters
and all insert their opinions. And I'm not saying that's
entirely wrong, because the Export Control Act provides that this should and must be the case. But all of
these opinions and decisions pile up on top of each
other with compounding-restrictiveness, I believe. It's
possible, of course, that the only way to get this really
clarified is for Congress to study the matter again and
come up with a revised or a new set of laws. And I
certainly don't blame the very competent administrators
in the administrative branches of the government who
are charged with the responsibility of enforcing these
laws for their overall-conscientious approach to the
problem. But the laws, I must say, might be interpreted
in any of several ways, and may be somewhat unclear.
About the most difficult problem that appears here
is, apparently, one that centers around a word that is
thrown around, called "strategic." I suppose there are
many interpretations you can put on this word. I don't
recall if it actually appears in the Export Control Act,
or some of the other Acts, but the interpretation is
always one of "strategic-for-military-purposes." This
appears to be of very great concern, particularly among
Defense Department people, and I'm sure that no one
in this room would deny the "possibility"; I certainly
won't.
It is mainly a matter of who is the prospective customer, and what he wants to use the computer for.
Most prospective customers are quite frank and open
regarding what they desire to use their computers forfor management information systems, industrial process
control, product engineering, etc. In this regard we
should rely on their representations with a reasonable
amount of faith, meanwhile oBserving through the customer liaisons open to the computer manufacturers,
that the computer is in fact being used as represented.
An item which appeared in Business Week, October
30, 1965, reported that COCOM regulations were recently modified to allow the shipment of nuclear reactors into the Eastern countries, provided that suffi-

cient controls are imposed to allow inspection which
would assure that those reactors were being used for
peaceful purposes. I presume that this inspection includes that by-products of these reactors (which can,
incidentally, be fuels for nuclear weapons) can also be
monitored. Now, if nuclear .reactors can be shipped,
with their potential for strategic weapons usage, I cannot see why computers cannot be shipped, because at
most, computers are only indirectly possibly useful or
"strategic" for military purposes. Computers are not
"weapons" per se. I· know even this point could be
debated, so I will get off it.
But, I say, there is a greater alternative possibility
here than we have tapped, and that it is the possible
"strategic" use of these computers for peace, to help
to politically orient these people toward our point of
view. I believe there is a very interesting problem here,
the answers to which hold tremendous opportunities
for the United States, and which can, and should, be
resolved in favor of the United States, vis-a-vis our
competitors.
MR. MAPES
Mr. Norman Ream is one of the best-known figures in the
field of systems research in this country today. He received
his Bachelor of Science degree at the University of Illinois and
is a licensed Certified Public Accountant in Illinois and California. Starting out in the petroleum business with Pure Oil,
h~ moved on to become Director of Accounting Research for
IBM. After a subsequent tour of duty with Lever Brothers,
he moved on in 1953 to Lockheed Aircraft Company, where
he served as Director of Systems Planning until, only a few
weeks ago he was appointed to his present position as Director
of the Center for Computer Science and Technology in the
Bureau of Standards. Mr. Ream is going to discuss problems
of marketing computers in Asia, especially in Japan, as well
as in some of the less developed countries. He also has a few
ideas on the educational aspect of overseas computer sales,
which I think he may share with us.

MR. NORMAN J. REAM
I think most of us are in favor of increased freedom
of trade with the rest of the world, and there are usually
two sides to every question, but Mr. Miles' suggestions
oil trading with the Communist countries move me to
express the personal opinion that we have to be very
careful about trading comput~r dollars for loss of
security. But that isn't what I came here to talk about
today.
Mr. Orr touched on the computer market in Japan.
My knowledge of that market has developed from a
series of lecture trips I have taken to Japan over the
last three or four years. I have visited Japan about
eight different times to lecture to the Japanese in-

The Overseas Computer Market
dustrial communities concerning the use of computers
in the United States and to discuss with them the
potential of their use in their country.
The Japanese computer industry is very interesting
in that today there are six major computer manufacturers active in the Japanese market. We have
American computer manufacturers who are also active
there-IBM, operating a large plant through their
Japanese company which is 96% owned, and Univac
being active in a joint venture company. Also there is
NCR, Burroughs, and CDC. Japanese companies that
are most active are Fuji Communication Apparatus
Company, Miksabichi Electric Manufacturing Company, which is part of the Miksabichi family, the Oki
Electric Industry Company, the Tokyo Shaboro Electric
Company and Hitachi Ltd. Hitachi is the largest company in Japan and has a very active computer manufacturing group.
The latest available figures I was able to secure date
back to September 1963, at which time there were
actively used in Japan 440 Japanese-manufactured computers and 285 American computers (of which 163
were IBM). Of course when you think that the first
computer was introduced in Japan in 1959, this figure
has now probably about doubled. But in increasing at
this rate the Japanese companies have gained a greater
percentage of the market than the American companies.
In other words the balance is swinging to the Japanese
manufacturer.
It is also interesting of course that Japan ranks second to the U.S. in the use of computers and I think
we will see a diminishing number of American computers in use in Japan in future years. I will try to tell
you why I feel this way by talking about some of the
problems in the development of the use of computers
in Japan. Their problems are not too dissimilar from
many of those we have here in the U.S. There is a great
shortage of skilled personnel. This they are attacking
through their universities, probably on a more formalized basis than we are. There is a direct emphasis
on this. Also recently there was a program introduced
in Japan by NOMA-the Nipon Office Management
Association; however, we shouldn't compare it to the
NOMA in the United States. It's quite a different type
of organization. They are embarking on quite a computer programming training course and they will probably have 20,000-30,000 students in this course within
the next 18 months. There also exists in some areas in
Japan the problem of management attitude that is a
sort of reluctance. However, I think that this reluctance
on the part of management in Japan is not nearly so
pronounced as it is here in the United States. They are

also faced with the problem of program language development. Most of the programming languages in use
in Japan are imported from the U.S., although there
are activities under way to develop their own languages.
Additionally, they also have some rather distinct
problems which are not as familiar to us. One is the
economic evaluation of the use of computers. Mr. Orr
touched upon this. They cannot justify computers solely
on the basis of replacing people. Currently their labor
costs are quite some bit below ours. But this problem
is changing in a very fast rising economy; however, it
will be quite some time before their rates will be comparable to ours. Consequently they are looking toward
the more sophisticated areas which Mr. Orr mentioned.
A second problem that they have is that they have
only nominal government support in the defense area
and admittedly in our country our government has
contributed very heavily to our defense efforts and all
our research and development efforts in the computer
industry. Japan also has minimal use of computers in
the scientific areas because most of their industries are
not heavily engaged in R&D activities, not as perhaps
many of us would like to think of it. We like to think
of Japan as being very active in R&D, and they are in
the electronic areas, but in many other areas they look
to licensing arrangements and the import of technical
know-how from outside of Japan. This is understandable, considering financial conditions.
They have another peculiar problem. The labor market is not nearly as fluid in Japan as it is here in the
United States. This is due to the paternal instincts which
are deeply imbedded in the Japanese industry, which
is sometimes referred to as "lifelong employment."
When a man or woman joins a Japanese company,
normally they stay with that company for the rest of
their business life. This means that in the introduction
of computers into a Japanese company they must start
at the very bottom and train their people to become
accomplished and acute. They cannot proselyte, as is
done here in Las Vegas, where there is sort of a trading
mart for personnel. I believe, however, that in spite of
all the problems they are faced with, they are going to
make a very extensive use and a very sophisticated use
of computers, especially in the management areas. I
believe that their management will react faster to
change, once the requirements for that change have
been determined.
Looking at the Japanese government support of the
computer industry, we find they are extremely active
and feel that the development of a very strong computer industry is basic in developing the well-rounded
electronics industry, for which they are very famous.

The Overseas Computer Market

In 1961 they established the Japan Electronic Computer Company, which is a leasing company to Japanese
industry. A special committee has also been set up
under the direction of the government. They are doing
extensive work in peripherals, and they are developing
equipment which will handle the common language,
which will ease the use of computers. One of the recent developments on the part of Hitachi has been the
electrostatic high-speed printer, which prints at 6,000
lines a minute. Two of these are currently in use in
Japan and of course they are developing printers which
will print the common language, this being one of these
developments. Their manufacturing techniques are extremely advanced and their quality assurance programs
are equally as good as any of those that are in existence
in the U.S.
While talking about some of the problems of the
American computer manufacturer, the Japanese government has for all practical purposes banned the import
of small and medium sized computers into Japan and
those that are imported from foreign sources have a
25% duty on the import. Of course this places the
foreign manufacturer at a considerable disadvantage
unless he is working directly with a Japanese company.
IBM has been manufacturing the 1401 and has been
granted permission by the Japanese government to
manufacture the 360-20 and 360-40. This means that
most American manufacturers in order to get into the
Japanese market are probably going to have to go
through licensing arrangements or joint venture companies with Japanese nationals.
There are several types of companies that can be
formed in Japan. One type is 100% American owned.
This is an extremely difficult task to accomplish and
one that is not used very extensively at this point. The
joint venture method is easier and one that is normally
used. In this situation we would find a Japanese control being exercised. In other words, they would own
more than 50%. There are many problems associated
with American companies working in Japan, i.e., the
problems of transport of technical. know-how into
Japan and the usually associated slow start-ups associated with a company until good working arrangements
can be made.
In Japan we are going to find that the Japanese government, while it is a democracy, is much more influential than we may suspect. We are going to find
that they will be very influential in supporting the use
of Japanese-developed computers in Japanese industry.
This does not mean that American computers will not
be used in certain areas, but I do think that the percentage of use of the American computers in Japan

will decrease in time.
As we look to the Orient in the lesser developed
areas, we have an entirely different situation. The
Japanese, American, and Western European manufacturers will be competing in a more or less open
market. Also there are many problems associated with
the use of computers in these less developed countries
and we are going to find, as Mr. Orr says, that these
are initially going to be used in the areas of government, or they are going to be used by the larger foreign
companies operating in these areas. I cannot foresee a
situation where local companies in Formosa, Thailand,
and Malaysia are going to have a need for large computers at this. time. This perhaps will develop over a
period of time, but I think the market here is much
more limited than we would like to realize.
I do believe, however, that in these areas there is a
great possibility for the American computer industry
to make a great contribution and to considerably improve the image of the United States. This is in the
area of education. Those of you who have visited these
countries recognize that there is a great void of middle
class people and that the industries are not going to
grow in these countries until the void is filled. It can
only be .filled through better education. I believe that
there can be an extensive market developed in these
countries through an educational field. However, this
is an area in which we have not done too well in the
United States. Perhaps by real emphasis in this area on
the part of computer manufacturers, the computer industries, and the educational institutions, a great contribution can be made.
MR. MAPES
Mr. Theodore L. Thau is the Executive Secretary for the
Advisory Committee on Export Policy. He went to the
University of Chicago, receiving a Bachelor of Philosophy and
a JD from the law school there and engaged in private
practice in Chicago and New York. Then he was Assistant
Solicitor for the Securities and Exchange Commission in
Washington. He spent a good many years as Assistant General
Counsel for the Department of Commerce in the field of export control until 1961, when he was appointed to his present
position. Ted has suggested that perhaps we could forgo his
initial presentation in the interest of opening a longer period
of time for questions. I don't think I can let him off the hook
that easily, but I can open the question by asking him to
give us a brief presentation concerning the parameters of the
term "strategic," which was discussed a few minutes ago, and
also by covering the special problem of the export of computers to France.

MR. THEODORE L. THAU

A few years ago, for someone like myself associated
with controls over exports from the United States to

The Overseas Computer Market
be a member of a panel concerned with overseas markets for anything would seem kind of phenomenal.
That's no longer the case and there is good reason why
I am here today-because the Export Control Act
which I'm concerned with is no longer regarded as a
complete bar to exports to the East European Communist countries, including the USSR, as it once may
have been considered.
As a preliminary matter, however, I want to tell
you-and this may seem a bit unfair, but I assure you
it had to be worked out this way in the interest of
brevity, among other things-Mr. Miles and I made an
agreement last night that if I didn't answer specifically
everything he said today this would not necessarily mean
I agreed with him.
First: Computers are licensed for export from the
United States to all free world countries, excepting
Canada, and to th~ East European Communist countries, including the USSR. I except Canada because
since World War II days we have had an arrangement
with that country whereby we do not require export
licenses for goods intended for use and consumption
there.
Our reasons for requiring licenses to export computers to the free world countries are not that we are
concerned with preventing them from getting computers
(with certain limited exceptions which I am going to
refer to a little later), but rather in order to prevent
unauthorized transshipments, re-exports, and diversions
from the free countries to the Communist world, including the East European Communist countries, the
Asiatic Communist countries, and Cuba.
We also require validated licenses fo rthe Communist
countries that I have described. Those countries really
no longer comprise one world, you know, but several
groups, toward which we have different levels of controls, reflecting the requirements of the law and policy
of the United States to safeguard oUI: security and welfare from those who might have hostile intentions
toward our country. This means that we do not deny
all applications for all licenses to all these different
groups of countries. We use the licensing technique,
instead, as the device to screen, from the strategic
point of view, orders from Communist countries for
computers. We do not grant any licenses for computers
to the Asiatic Communist countries, for strategic and
foreign policy reasons. I will not go into that any
further here today. The same applies to Cuba.
With respect to the East European Communist countries, our restrictions are more selective. You will notice
that I have been careful not to refer to the Soviet bloc ..
The reason is that we no longer regard all of the East
European countries and the USSR as constituting a

bloc. Just a couple of years ago we came to the conclusion that it no longer made sense to refer to anything
called the Sino-Soviet bloc. So, now we consider that
the East European Communist countries and the USSR
are all to be treated individually under our export
policies. For some of those countries, this will mean
more favorable treatment than others. For example,
since 1958 we have treated Poland more favorably.
We will even allow her to receive strategic goods, if
they ate found to be reasonable and necessary to the
Polish civilian economy. In the summer of 1964 we
entered into negotiations with Rumania, as a result of
which we gave Rumania a preferred status, much as
we have for Poland. In accord with that preferred
status, we allow Rumania to receive certain strategic
items, if they can be found necessary and reasonabl~
to the Rumanian civilian economy. Each of the other
East European countries must be looked at on its own
footing, in the light of our foreign policy and strategic
interests relating to the particular country.
What do we mean when we say that we use export
controls to screen the strategic from the nonstrategic?
If we merely were to limit ourselves to direct military
items, and be concerned only with tanks, guns, planes,
bombs,etc., that would be one thing. There are controls
over items of those kinds maintained by the Department
of State Office of Munitions Control. Some of you may
have had experiences with that office in connection
with applications to export to free world countries
items which you know or have reason to believe are
going to be used there for military purposes. However,
the U.S. Export Control Act, which is administered by
the Commerce Department, goes beyond the narrow
concept of direct military use. It embraces a concept
that might roughly be called a "military-industrial
mobilization base." Some phrase of that sort would be
aptly descriptive of the area of concern that is involved when the Commerce Department uses the term
"strategic. "
From that point of view, then-using that concept
as our yardstick-we have in the past approved licenses
to export some kinds of computers to Communist countries. We approve them now to the East European
Communist countries. We also approve components for
computers which you may send to your West European
affiliates or other free world firms to be used in the
making of computers to go to East European Communist countries. We also approve peripheral equipment to go into computers to be made abroad and sold
to the East European Communist countries. There is
no U.S. embargo on all computers, components, and
peripherals for the East European Communist countries.
There has been some misunderstanding about this. It

The Overseas Computer Market

may have been because, in addition to concern about
building up the military potential of the East European
Communist countries, the Congress required us in 1962
to be also concerned about contributing to the buildup
of the economic potential of the East European Communist countries, However, this does not mean, has
never been interpreted to mean, that every computer,
because it contributes inherently to the economic potential of an East European Communist country, is also
necessarily detrimental to our security and welfare. We
have publicly interpreted this amendment as not requiring us to conclude that an item is detrimental to our
security and welfare, from the economic potential
standpoint, if a comparable item is readily available to
Eastern Europe from other free world countries. We
have, therefore, a basis on which, even applying the
economic potential criterion, we can and do approve
licenses to export computers, computer components,
and peripheral equipment to East European Communist countries.
I turn now to the kinds of computers, components,
and peripheral equipment, that we deny, that we are
concerned about, that we ask you many questions about,
when you come in to us with your license applications.
When you tell us that you are interested in trying to
sell these, we ask you, "How advanced is this computer over what is available from the free world without the use of U.S. components, without U.S. technology?" "How advanced is this over any that would
be available to the East European Communist countries
from their own resources?" We are here concerned with
the more advanced types of computers. We are concerned with these because we have been told by the
technical experts from Defense and other government
agencies, that it is in this advanced area that the United
States has a significant lead over the East European
Communist countries and that there is more likelihood
that such computers will be used in part and substantially for military-industrial, mobilization-base activities
than the less advanced types of computers.
You may argue that this isn't necessarily so, and
that is quite true. One could use the most advanced
computer in the world for peaceful purposes. If we
could be certain about that, it would be fine. I believe,
however, that one would have a good deal of difficulty
in finding a basis for even a reasonable degree of
probability of such limited usage. For example, if company X gets an order for one of the most advanced
types of computers presently known from a department
store chain in one of the East European Communist
countries, we may find ourselves having to ask the
question: How many hours a week will this computer
be needed to do the work of the chain? It may turn out

to be that only a few hours a week are needed. What
will be done then with that very expensive monster
the rest of the week? We know, and it was a thrill for
me to see this kind of operation yesterday, how peripheral equipment hundreds of miles away from a giant
computer can be so connected to it as to work out very
advanced problems, and how many pieces of peripheral
equipment can be connected up and operating at the
same time and still using only a small fraction of the
capabilities of this computer.
Now it doesn't follow, even from what I have said,
that there is anything that requires denial on the basis
of this or some other single factor. We need full information from you in those cases. We need specific
knowledge to be able to meet the requirements of the
law and the national policy. We need to be more sure
with respect to those types of computers than with
respect to the others.
Next, it isn't enough just to tell us that "comparable"
technology is available from computers that are obtainable from West European countries. The technology
in a wide range of computers may be comparable, as
technology goes. But the computers themselves may
vary quite widely, and some may be far more advanced
than others, even though all use comparable technology.
Of course, we are concerned to prevent our advanced technology from going without authorization to
the East European Communist countries. We are also
concerned not to have our most advanced computers
go. As one of the people here at this conference told me
yesterday, t1)e East European Communist countries recognize that by getting our most advanced computers
they have the best chance of overcoming most swiftly
our technological lead. We have been queried whether
this technological lead they want to overcome is only
in the economic area, not related to a military-industrial
mobilization base. Well, we're not sure. We're therefore concerned. So we will ask you many questions
when you come in with such cases.
I would now like to talk a bit about one problem
area, outside the East European Communist area, with
which some of you have had experience. I refer to the
problem that has arisen as a result of the Nuclear Test
Ban Treaty, and especially as a result of certain activities of a friendly country that is not a signatory of
the Nuclear Test Ban Treaty. We have obligations
under the Test Ban Treaty not to aid any country in
the development of nuclear weapons or in conducting
nuclear explosions and nuclear weapons tests. As a
result, the Commerce Department and the State Department's Office of Munitions Control have adopted
regulations of a complementary nature, designed to
regulate by special licensing procedures, items that /

The Overseas Computer Market

may be used for such purposes, even if going to free
world countries, and whether or not specifically designed or modified for use in nuclear weapons testing. As some of you may know, we have expressed
considerable concern here about very advanced com-

puters, components, and peripheral equipment.
There is much more I could tell you about this new
area of export controls, but time is running out and
perhaps in some of your questions you may ask something about it.

A Panel Discussion

The future of electromechanical mass storage

Chairman:
WILLIAM A. FARRAND, Autonetics
Representatives of Large Installations:
ROBERT M. FRANKLIN, Chrysler Corporation
NORMAN HARDY, Lawrence Radiation Laboratories
ROBERT M. GRAHAM, Project MAC at Massachusetts Institute of Technology
MARVIN EYSTER, Woodward Governor Company
Equipment Manufacturers:
WILLIAM J. BRODERICK, General Electric Company
FRANK J. LOHAN, Bryant Computer Products
DONALD K. SAMPSON, Control Data Corporation
ALAN F. SHUGART" IBM Corporation
IRVING L. WIESELMAN, Data Products Corporation

INTRODUCTION *

frame of reference.
In order to address ourselves to the future of mass
stores, it is necessary to determine their present position. From their shortcomings and inefficiencies, it may
be possible to determine their immediate path. The
distant future (over 10 years off) is nearly impossible
to forecast and is very dangerous for we'll probably live
to find our predictions false. It is our intent to discuss
the "needs and the possibilities" of certain specific
characteristics. By having system users detail their
problems and equipment manufacturers discuss their
solutions, it is possible to see the cost and utility of
equipment and, therefore, to judge the worthwhileness.

The field of mass storage under consideration by the
panel is that which communicates at machine speed in
machine form. The areas of concern are both software
and hardware from total system to details within the
components.
The amount of mass storage presently installed in
the U.S. is above 1015 bits in an on-line or pseudo online manner. Some organizations have made sincere
requests for this much data capacity in a single store.
Over eight years ago Art Angel described the perfect
auxiliary store as an "infinite size memory instantaneously accessible and economically feasible." Ideals like
this are seldom met, but the future lies between the
present and the ideal. The concern for the future is not
what can be out there but rather what is going to be
out there. What determines the future is present and
future problems in light of the technical and economical
feasibility of them. The future, when you get there,
should be the past-not still the future. This is our

THE CHRYSLER 50,000-MILE
WARRANTY (RACE) SYSTEMt
The Chrysler Corporation has 40 to 50 random
access systems in operation. The 1.2 billion-character
on-line RCA 3301-RACE system located in Highland
Park, Michigan, is used to support the Chrysler five-

t Robert M. Franklin, General Supervisor of Information
Systems Planning and Research for Chrysler, reported on the
Chrysler system (summarized).

*William A. Farrand, Assistant to the Chief Engineer, Data
Systems Division, Autonetics, set the stage for the session
(summarized).

The Future of Electromechanical Mass Storage

year, 50,000-mile warranty program. It is connected
via commercial communication lines to video data
terminals installed in regional offices in 23 states
around the United States. Each of the 14-inch video
data screen and teletype keyboard terminals is capable
of getting into the file to find the history on a vehicle.
Within 2 Yz to 3 seconds after a request of a vehicle
by serial number, the operation has up to 480 characters
of information displayed on the history of that vehicle
from the day it was manufactured. A page-turning
operation is available which permits request of additional information in· a particular vehicle file when 480
characters is insufficient for the complete history.
There are 1.2 billion characters on file now, and the
program is only three years old. A firm requirement
exists for 9 billion characters of on-line storage. An
order of magnitude increase from this size is foreseen
on this application alone. Each RACE stores 400 million characters on a stock of 256 random access magnetic cards with an access time (random average) of
333 milliseconds. With an increase to 9 billion characters, this average access time will not increase. This
access rate poses no problems for the remote inquiry
system but poses considerable difficulty when updating
the file (present updating time, 20 hours, caused in part
by the large block size-650 characters). Quality Control and Engineering people at Chrysler are beginning
to appreciate the data which are stored in the file, but
present organization of the file makes it nearly impossible to extract information to fill their requests as
the file was structured to service the remote inquiry
stations.
The file was installed in F~bruary with very few
problems which were not solved by component selection. The commanding reason for choosing this particular equipment for this particular installation was
file cost. At the selection time this was one-third its
nearest competitor.
THE LAWRENCE RADIATION
LABORATORY COMPLEX*
Current equipment includes two 7094's, two 3600's,
a LARC, a STRETCH, and a 6600. An additional
6600 is expected in April and a 6800 in early 1968.
Current accessory equipment includes three large Bryant
disk files with the 6600, a 30,000 line-per-minute
printer, miscellaneous input/output stations for individual researchers' use, and the 12 drums associated
with LARC. An IBM 2321 with 3X109 bits capacity is
*Norman Hardy of Lawrence Radiation Laboratories outlined the Livermore installation and the operations contemplated (summarized).

expected momentarily and a 10l2-bit (equivalent to
approximately 10,000 reels of high density tape) storage unit is on order from IBM. In general, the operation
of the mass storage elements is repetitious in nature
and requires only simple programming to permit simultaneous reading, writing, and computing on different
elements. Operations tend to be run in a time-sharing
mode during the daytime with long physics problems
handled at night when personnel demands are not so
great.
At the present time the computers and mass storage
equipment is scattered in a number of adjacent buildings. It is planned to move them into a single large
room. Ultimately it is planned to have teletype units
and TV-type display systems scattered widely over the
facility for the convenience of scientific personnel. The
majority of these will be within a thousand feet of the
computer center. Thus the entire operation is comparatively compact relative to most time-shared systems.
It was noted that the special printer produces about
10 pages a second and eats up several hundred dollars'
worth of paper an hour. Consequently, one of the objectives of the organizational planning is to provide
more direct access to computer output so that the need
for large-scale printout can be reduced. At the present
time one microsecond memory having a capacity of
128,000 36-bit words is available connected to six
independent memory busses. One of these busses will
be used solely by the General Precision disk file which
has a 10-megacycle bit rate and 8X lOS-bit capacity.
This disk file consists of two units, each having 5,000
fixed heads, of which nine are active at one time. The
interconnecting channels currently operate at the 10
megacycle bit rate, but Norm Hardy considers this extremely slow for use with machines such as the 6800
and anticipates that in the 1970's equipment will be
available for operation up to 1,000 megacycles.
A PDP-6 unit is used at Livermore exclusively as an
executive controller to allocate work to the other units.
To obtain maximum efficiency, allocation of disk storage space is crucial. Current thinking is to allocate disk
space by pages of 32 sectors of 1,024 words, separated
by one-sector gaps. Pages will be identifed by one of
18 standard page angles, and tasks will be queued by
page angle. When none of the 18 queues are empty,
95% of theoretical transfer rate will be possible (equivalent to 250 accesses per second if each page is considered an independent access). Methods allow for
accessing any arbitrary page or pages from a complete
data file without following the entire file through
serially. It is expected that the major data flow will be
between the disk and the other devices in the system.

The Future of Electromechanical Mass Storage
PROJECT MAC*

seeable future. The 720-A and 1302 used on the MAC
system are both underpowered, as the former has a
quarter-second 32,000-word swap time, aQd the latter
has effectively only a single access path. For such
service this kind of equipment needs a multiplicity of
access paths in particular in order to be of maximum
utility for a service center. Obviously the faster the
access and transfer times, the better, as it approaches
more nearly the capability of an all-core memory.

The MAC system is basically the first prototype
operation of a computer utility on a 24-hour basis. The
next version, described elsewhere in this conference, is
called MULTICS. The basic objective of these facilities
is to provide service to a large number of terminals
scattered over a large geographic area. This service effectively permits independent use of the central computer and storage facility by each terminal and permits
each terminal to expect that all information related to
its activities would be retained in the facility memory
until explicitly deleted by that terminal. At the present
time Project MAC uses a 7094 computer to service
approximately 160 terminals in the New England area
of which 30 may be active at anyone time. For the
type of operation contemplated, the ideal memory is
obviously unattainable since it would have infinite capacity and instantaneous access and transfer at zero
cost. The compromise approach used to approximate
this in the M.1. T. program is to provide a hierarchy of
existing types of memory of decreasing speed and increasing capacity within the limitations of available
funding and space. Thus, the highest-speed units are
core memory followed by drums, disks, and tapes in
the present system. It is conceivable that some detachable memory may replace the tape units in the future.
In addition to the 7094, Project MAC currently is
equipped with one 1302-disk unit, two 720-A drums
and one 720. Programs are given core space sequentially, with at most one user control program ready for
execution at a time. The general intent of the facility
is to develop operating and programming routines to
relieve the user at the terminal from any consideration
of the type of auxiliary memory equipment with which
he is operating and to relieve the management of the
detailed format of files at the Center. Thus the Computer Center is undertaking to handle the interconnection between terminals and the actual operating equipment in such a fashion that Center equipment can be
changed from time to time without influencing the
user's methods of operation. In terms of thinking on
the management of the hierarchical memory, it is
planned that programs will be transferred to slower
and slower storage media when they are inactive, and
that active programs will be maintained at the highest
level of accessibility consistent with system capabilities
and overall user demands. Present thinking is that highspeed drums and/or disks will continue to be the major
units required for readily accessible storage in the fore-

This company has a rather specialized requirement
since they produce and maintain a stock of some 55,000
component elements from which they custom-assemble
governors to meet the requirements specified in customers' orders. To provide inventory control, update
engineering specifications, coordinate requirements of
customer orders, schedule the use of machine tools, and
consolidate cost accounting work with, the accounting
department, this company makes use of an NCR 315
CRAM system. This includes a 10K core memory, four
CRAM units, and equipment for reading and punching
cards and paper tapes. This equipment took over for
prior punched card equipment for the same applications. In addition to the manufacturing control operations, some forecasting and proiecting of requirements
for individual components and generation of appropriate manufacturing orders is carried on. By using
random order in file organization, the access time in
general is about 235 milliseconds, and in the special
case where information happens to be on the card on
capstan, the access time is about 45 milliseconds. At
the present time approximately 88 % of the total capacity of the file is in use and operating in good order. All
records in the system are stored in a single, randomordered file. All accesses to the file are made for the
operating programs by a common executive program.
Since the CRAM cartridges are removable, duplicate
copies have been made for use during debugging operations to prevent danger of destroying the good files of
information by programmer error. Improvements which
Marvin would like to see in equipment include faster
access time and a write lock-out on the file under program control, but only if the improvements can be
made with complete compatibility between the improved devices and the current ones to avoid reprogramming difficulties.

*Robert M. Graham, Program Coordinator at MAC Center
of M.I.T., reported on the Center's system and plans (summarized).

*Marvin Eyster, Manager of Data Processing for Woodward
Governor Company of Rockford, Illinois, described their use
of data processing and storage equipment (sumfilarized).

THE WOODWARD
GOVERNOR MANUFACTURING
CONTROL WITH CRAM*

The Future of Electromechanical Mass Storage

QUESTIONS AND COMMENTS
FROM THE PANEL*
Bob Franklin pleaded for techniques to extract information from a file based on requirements not defined when the file is originally structured, and he
postulated that this was a hardware problem. Irving L.
Wieselmant answered that he considered this a software problem. The hardware manufacturer is concerned with building better hardware. More capacity,
faster transfer rates, and lower costs are his goal. Bob
Graham supported the position that this is a software
problem, and pleaded for general-purpose programs
which are hardware-independent. He further strengthened his belief that all the special wrinkles should be
in the software and that the hardware should be as
general purpose as possible.
Franklin's rebuttal brought out his disbelief that the
electromechanics of mass files are not strongly influenced by data structures (e.g., the rigid formats
characteristic of most present machines). Don Sampson:!: stated his belief that manufacturers build what
sells, and overly specialized equipment does not sell.
Graham pointed up the difficulty when the difference
between the logical structure of the data and the physical structure of its representation are confused. He
contended that the two are completely independent but
are all too often made dependent and that this makes
a special-purpose device. One of the objectives of the
MULTICS system is to keep the users' logical data
structure completely divorced from the physical representation. This facilitates the updating of hardware.
Bob Franklin displayed a RCA RACE card and
pointed up the wear problems with it. There was no
apparent wear on the magnetic recording surface. The
wear showed up in the ejection area and on the picker
studs.
Wieselman noted that the panel represents users
which require exceedingly large data bases, and he requested comment on the value of a very small file used
for a small data base for a small business. His question concerned the usefulness to a small business of
privately-owned EDP vs public utility EDP.
Farrand noted that there are a very, very large number of disk packs around and the ability to get the data
and store it on these seems to be a very well accepted
practice.
Someone not identified from the floor suggested that
address by content is the answer to Bob Franklin's
*This is a summary of a tape of the panel's actual comments.
tVice President of Product Management, Data Products
Corporation.
:j:Director of Engineering, Mass Memory Operations, Control
Data Corporation.

query. Bob agreed that updating of their file is an
address by content operation but that the method is
exceedingly inefficient for they must "page" the data
into core to address by content, and for the 1.2 billion
characters in their system, this takes 20 hours. Additional data were revealed at this point concerning the
Chrysler file. It has 6 million vehicle records and each
record is quite large. This needs to be a large record
for, in just the Plymouth division, there are 13 million
buildable models. He then pleaded that there must be
some way of looking right into the information within
a random access file and saying, "This is the way it
should be," and out it comes-even though it was not
set up to have this request made in its original formatting. Bill Farrand noted that there is a dearth of algorithms concerned with adaptive storage but that address and extract by certain tags is quite another story
and is quite practical in some cases. He further commented that at present this is circuitry-wise feasible in
small files and is truly useful in huge ones.
QUESTIONS AND COMMENTS FROM
THE AUDIENCE AND PANEL
Byron Smith * directed a question to Alan F. Shugartt concerning IBM's interest in devices of the
CRAM and RACE character and made specific reference to the 2314, sometimes called a wall-to-wall disk
file. He asked a specific question: ""Is the access time
of the card file a basic detriment, and is that why you
are not using it?" Shugart answered that IBM is in
the card file business with their Data Cell Drive. Shugart acknowledged the competition between the 2314
and the 2321 and stated that he thought there was a
reasonable chance that the 2314 might replace the
large disk files but not the Data Cell. Al then showed
a movie of the first installation of a 2321 Data Cell
Drive outside of IBM, a film on the installation of their
machine at Allstate Insurance Company. This is used
in conjunction with a Model 40 system. This installation has two 2311 disk packs, a 2321 Data Cell Drive,
together with a 2841 control unit. These films showed
the Data Cell Drive being installed and running. The
Data Cell Drive was shown in operation and an operator
was shown removing cells from the drive unit and reinserting same. An unidentified voice from the audience
asked the question concerning expected life in terms of
number of passes of a card in tape strip files. Shugart
stated that they do not track the life of their cards in
numbers of passes but that their specifications call for
4 to 5 million picks. He emphasized that this is not
*Analex.
tTechnical Manager, Random Access Memory Programs,
IBM Corporation.

The Future of Electromechanical Mass Storage
necessarily addresses b.ecause once you pick a strip,
you may have several addresses on that same strip.
And he emphasized that 4 or 5 million picks per data
cell should be the life of that cell but that revolution
life around the capstan is not expected to be a limiting
factor. Marvin Eyster commented that their installation
seems to get about 200,000 revolutions on the drum
per CRAM card but that they seldom revolve a card
more than one or two revolutions and card life is
limited by the number of times they drop it, and they
are certainly experiencing in excess of 10,000 drops
per card. However, they do not keep log sheets in this
manner on the data.
Bob Franklin noted that the card stock on the RACE
units is manufactured by 3M and that they have made
some significant improvements in the material. He
further noted that oxide life was not a determinant on
the card life. William J. Broderick* commented on
Byron Smith's point concerning the multitude of seeks
available with a 2314 disk complex when used in
parallel seeking, and noted that this could effectively
reduce access time. He then asked the questions: "What
does overlap access mean to us? What is the value of
parallel read/write? What would it mean for the
Chrysler RACE system to have multiple read/write?"
He further asked Franklin if they have parallel read/
write control. Bob answered by stating that they are
not able to operate all four units simultaneously.
Frank J. Lohant mentioned that there are several
pieces of equipment on the market with simultaneous
read and write but that the File Drum is the only current
piece of equipment on the market in which several independent actions can be taking pl~ce in a given band
of data simultaneously. Don Dittberner:!: expressed concern over the panel's apparent emphasis on present
systems and lack of emphasis on future systems. He
passed out a query concerning what is detrimental to
having more control associated directly with disks so
that the experience with serial associative search mechanisms could be utilized. A second query concerned
the feasibility of a single unit incorporating a combination of optical recording and magnetic recording. Also
why hadn't the question of packing density been brought
up? He tendered the belief that 3,000 to 5,000 bits
per lineal inch isn't going to be an unusual density
in the near future. Bob Franklin remarked that the
four users on the panel had pointed out rather graphically their problem areas and what they will demand
in the near future. He emphasized that multiple access,
high speed, high reliability, and low cost are certainly
*Product Planning, General Electric Company.
tDisc File Product Manager, Bryant Computer Products.
:!:Center for Management Technology.

in our immediate future. Bill Farrand commented that
the point brought out by Don Dittberner concerning the
possibility of making the file a most autonomous entity
by having closely associated with it certain logical editing and manipulative functions is a very good one for
some people but is in definite deference to the attitude
portrayed by Bob Graham. It was further mentioned
that the 6600 and 6800 systems were structured in just
this manner with peripheral control.
Bill Farrand then pointed up the controls on the
future as being economic value and technical reliability.
Files can be structured so that they do not take up
main frame master control program steps. But are these
added features worth the price? Next, reference was
made to the 1012-bit file ordered by LRL. As the large
present disk files are about 1 billion bits in max size
and larger than 4 feet in case length, using the electromechanics from these to form a 1012 file takes 1,000
units or over 4,000 feet. This demand for nearly a
lineal mile of floor space might make the unit unfeasible.
The exchanges which we hoped for between the
panel and the audience were thought to temper the
question, "What will be available in the future?" by
answering questions beginning, "My systems needs are
. . ." by: "The following technology will answer this
with the following characteristics at the following cost.
Are you willing to pay for it? If so, we'll build it."
The questions are, where are we going, what are we
going to have when we get there, and what do we
really need if we're to get there. The only way to get
hardware in use is to have specifications on equipment
which can be met by the manufacturers and which are
useful to a large number of users.
Alan Spahr* asked for Al Shugart's opinion on the
place of fixed vs removable media in disk files, and Al
answered by stating that it is his opinion that a replaceable file is here to stay. He reinforced his opinion with
considerations of reliability, noting that you can move
the storage media from one machine to another if the
machine breaks down. Don Sampson made the distinction between, removable disk files and large fixed
disk files by calling the fixed files large precise disk
files. He further expressed the belief that there is a
definite reliability edge in favor of the precise disk file.
His belief is that the disk file business is going to settle
down and that the bit density and cost per bit will
decrease drastically on precise disk files. He, however,
emphasized that though the race is on between the two
types of units, neither will die for want of customers.
Bill Broderick requested some discussion concerning
*Analex.

The Future of Electromechanical Mass Storage

the reliability difference between flexible media and
media in storage subsystems. ~e specifically inquired
if there was anything inherently better in flexible media
other than price over fixed or removable disk units. Al
Shugart answered by stating that the 2321 was developed strictly on a cost-per-bit basis. Bill Farrand then
made sqme comments concerning the excellent volumetric efficiency of a strip file at least in principle.
He also alluded to the value of the removable media
feature usually built into strip files. Bill Broderick
wondered if the use price to the customer were the
same for a mass storage device that did not employ flexible media but was removable at the same
price per character, would not the operational characteristics pay for the flexible media? Irv Wieselman
summed up by asking if Broderick's real question was,
"Are disk files more reliable than strip files?" Al
Shugart answered that in his opinion disk files are more
reliable than strip files. Al reinforced this opinion by
stating that we have a large amount of experience with
fixed media files and that this position may change
when we have comparable experience with flexible
media files.
Farrand commented that the removable media files
would always be plagued with dust and dirt to a greater
extent than fixed media files. An unidentified person
from the audience then questioned whether removable
media and fixed media files could be compared, since,
in the hierarchy of stores as he saw it, they were on
different levels, and his point was we were comparing
apples with oranges.
Bill Farrand then commented that our experience in
strip files isn't as sparse as one might think for nearly
a thousand such files exist today. Bill Broderick then
showed a slide, which gave cost per character (normalized to six-bit characters) vs record accesses per second
for various types of files (Fig. 1). The prices depicted
are from total subsystem on-line. The systems have
been categorized by function: A, being core extensions;
B, recall processing; and, C, very large capacity reference file units. The dotted outlined area A is "announced future large solid-state storage." The solid
outlined area A consists of drums and fixed head disk

files. The pricing data predictions are obtained from
AFIPS and IFIPS publications.
SUMMARY *
What light have we thrown on the equipment and
systems we have with respect to ordering of data, interrecord gaps, deskewing, and structure? Is the choice
proper-as often used-of taking parallel data, turning the corner with it, and streaming? Is it an error to
not stream the data more since, on the average, this
is such an efficient operation for long records? We
have attempted to state that if customers need it and
technology can produce it, "it will be," if it is economically feasible.
The view taken of the future is what will be-not
what might be. If the needs of the industry are distinct,
individual, all specifically different, then the economic
feasibility is to be questioned. But if the need is universal and technically feasible, then the need will be
answered by equipment. This is only prevented when
poorly designed equipment is installed in systems and
their failure to work costs so much in time and money
that the situation is leapfrogged.
Many believe that the perfect organization is similar
to the way a random access core is organized. This is
often not an optimum organization due to the difficulties of locating the data by programming. The
block organization of data in many electromechanical
memories is of significance in many installations as is
the sequential flow of data in a record. Electromechanical mass memories of many kinds and forms exist with
excellent up-times and very low installation cost.
The questions are: "What is really wanted? What is
really feasible? How much does it cost? Does that make
it unfeasible?" Technology is leaping forward. One
characteristic brought out by the panel members is that
software is such a large investment that it is absolutely
necessary that any future system be able to work-no
matter how crudely, no matter how poorly, no matter
how inefficiently-with the data in the available form
using the available bases.

*By

the session chairman (summarized).

MILS PER CHARACTER OF ON-LINE STORAGE
PRICE PER CHARACTER (NORMALIZED TO 6 BITS): Purchase price of controller plus first
full storage unit, divided by formatted capacity in 6-bit characters.

1.000

~~~.

500

rtf ~

400

300
200

A~ ~

100

o LIBRA CO

fl\ tn-..",

1~
-

t'"

•

~M ""VI

.-"v

. V•

5.0

I'''

o IliM

3.0

~/J'J'J'J'J'J'''
'HHHH",

.03

PE OOP

T~!~~ON-L!~

'///////////)

II'tvuu

~
~

If///fl///IJ:

oJBM IDA'
0

:Js~E
---

.02

0.1
·0.01

DI\'''I

( IBM 13 2
0
E05-25

If/HO'NUj

0.5
.04

POTTER

~Hl

o RR ASTRAND
IV

rJ/'~:':...... ::'J'~

1.0

c
c
0

-'0 IBM 131p

".0
2.0

lL!:t

BU RO GHS

~
108 ~m
~,
I
I-r- I-,

CR'fi LECTRIC
rv"EM( RY 197
10 BITS

••

o E~LK CORI

~~E~: I~G ~~~~

E EC RON ~EAM
01 piNG 1968
7
RI lb BI S

-UU\
~

o IBM 131
o DATI -DISC

PIY'L"U

i~
iiiii

o I M 7320

.02

.03.04

0.1

0.2

0.3 0."

3.0 4.0

30 40

100

200

300 400

1,000

2,000 3,000 4,000

10,000

ACCESSES PER SECOND: Data-dependent (non-simuttaneous)access rate determined by average
positioning time (if any) plus 1/2 latency plus full latency (minimum to initiate an update).
This time in mi II i seconds is then divided into 1.0 second to determine a "minimum access rote~'
potential of the subsystem.

100,000

1,000,00l

William J. Broderick
General Electric Company
Phoenix, Arizona
1965

j•.
~

a.
~

Figure I

A Panel Discussion

Promising avenues for computer research
REX RICE, Fairchild Semiconductor, Palo Alto, California
KEITH UNCAPHER, The RAND Corporation, Santa Monica, California
TOM STEEL, System Development Corporation, Santa Monica, California
L. C. HOBBS, Hobbs Associates, Corona del Mar, California

PART I

FUTURE COMPONENTS
Rex Rice

Part I of the session is a presentation by the four
panelists. We would like to give you our views on computing as it can be in the year 1970. In Part II, we
invite the audience to comment. We'll be very pleased
to have you air controversial views. We on the panel
agree we do not know all the answers. First, we want
to isolate promising avenues of research related to the
potential payoff, secondly, we want to emphasize either
existing research that should be encouraged or research that is not now active and is needed.

INTEGRATED
CIRCUITS

SYSTEMS GENERATION
Figure 1 is a simplified presentation of two generations of research, development, and manufacturing
cycles of data processing systems. We are now in the
production phase of the integrated circuit generation.
A Jarge amount of research effort is presently being
expended to develop large batch fabricated arrays of
components. As shown, we are now in the early re-

~j@rrrrr{f{~:.Q{tJtmmmml ~ -:._______-~ 3

USE

INTEGRATED
MICROSUBSYSTEMS

Figure 1.

Promising avenues for computer research

ONE ISOLATED TRANSISTOR
SEVERAL DIODES
SEVERAL KILOHMS

DEVICES

.002"1: :I

T-:t

t.-.002

500 / LINEAR INCH
250,000 / SQ. INCH

MULTILAYER METALLIZATION

•

FLIP SILICON MOUNTING

•

HIGH YIELDS ----. ON DIES .250" PER SIDE

•

CONDUCTIVE COOLING
Figure 2.

search and development phases of the "Microsubsystern" generation and can anticipate such systems as
being in operational use by 1969 or 1970. The balance of the discussion herein is concerned with the
next generation and not our present "integrated circuit"
generation.
SILICON TECHNOLOGY AND ECONOMICS
The rapid advances in integrated circuit technology
and manufacturing capabilities have already outdistanced the predictions made several years ago. As
illustrated by Fig. 2, we are now making devices which
are only 2 mils on a side. Within this area one can
obtain a transistor, several diodes, or several kilohms of
resistance. This density represents 500 components in
a linear inch. It also provides the amazing number of
250,000 components in a square inch.
This component density when considered alone, has
no meaning. It must be coupled with other processes
presently being developed in several research laboratories. First, multilayer metalization must be provided
over the top of these devices in order to interconnect
them into a useful microsubsystem. Secondly, a desirable but not mandatory process is the flip (inverted)
mounting of silicon. Thirdly, research and manufacturing development is being vigorously pursued to obtain
large devices at high yields. One can anticipate that a
substantial improvement in yield will be obtained in
the near future. We can then seriously consider a highyield device which has a dimension of a quarter of an

inch on a side. Finally, techniques are already available which allow heat developed in the silicon to be
conducted out and dissipated elsewhere (see G. E.
Moore, "Cramming More Components onto Integrated
Circuits," Electronics, Apr. 1965).
The economics of micro-systems resulting from this
research can be summarized as follows:

Circuit Assumptions:
65,000 devices per quarter sq. inch die
20 devices per circuit
storage bit
flip-flop
gate, etc.
3,250 circuits per die
Cost Assumptions:
$50 silicon
10 package
40 testing
$100 for 3,250 circuits
Less than 3c per circuit.
There are approximately 65,000 devices in one quarter-inch die. Let us assume the average of 20 devices
for each circuit. Each circuit could represent a storage
bit, a flip-flop, a multiple input gate, etc. This provides
3,250 circuits on a die. For high-volume production on
a few types of dies, one may assume silicon costs approximating $50, packaging $10 and testing of completed components $40. This totals approximately $100

Promising avenues for computer research

8'7

1,000 LOGIC: 10 MICROSUBSYSTEMS
200 MULTILAYER P.C. BOARD

I 1,200 FOR 32,500 CIRCUITS
Figure 3.

for 3,250 circuits. It is technically feasible to achieve
high-speed integrated microfunctions with circuits approximating 3 cents per circuit by 1970. It should be
emphasized this will require some restraint on the part
of circuit and logic designers so we can develop standard microsubsystems to achieve high-volume production
on each type.
In Fig. 3 a hypothetical system function is shown.
This function contains 10 microsubsystems mounted on
a multilayer printed circuit board. Assuming a $200
cost for the multilayer circuit board, the total function
cost would be $1200 for 32,500 circuits. For example,
this single card could represent 500 words of 62 bits
of high-speed storage. Alternatively, it represents more
logic than is found in 7094 central process unit and
its instruction controls.
BULK MEMORY ECONOMICS
Bulk memory will undoubtedly take many forms by
1970. For the purpose of this discussion, it is only
necessary to illustrate one bulk memory with a reasonable cost so desk-side, real-time, stand-alone computing may be performed. A simple extrapolation on disc
pack type memories now in development, as discussed
at the Batch Fabrication Conference, follows:

Development now:
Discs-12.5 X 106 characters on-line
20 accesses per second
10,000 characters per second flow
$12,500 drive
500 pack
$13,000
Research now:
BORAM-4 X 106 characters on-line
1 microsecond access
1.5 X 106 characters per second flow
$???? read-write unit
500 pack
$????
It was stated that for $13,000 the system would
provide 12,500,000 characters on-line. At 20 accesses
per second and 500 characters for an average access,
this system provides a data flow rate of 10,000 characters per second. This flow rate will be adequate for
some purposes; however, we should anticipate that with
proper research higher flow rates will be achieved.
At the Batch Fabrication Conference, BORAM was
described. BORAM is a block oriented random access
memory of an electronic nature. At the present time a
research program is funded to do work on this system.
The specifications call for 4 million characters on-line

Promising avenues for computer research

in a removable pack which costs no more than $500.00.
The specifications also call for I-microsecond access
time and a 1,500,000 plus character per second flow
rate. Systems designers certainly hope that this type of
memory will be developed. The two memories discussed
here illustrate what can surely be expected and what
we may hope to achieve by 1970 with proper research.
Hypothetical System
A hypothetical system, utilizing components previously described, i~ presented in Fig. 4. Here we see a
desk which contains the logic and disc pack storage
drive in the drawers on the left. On top of the desk is

storage. This Memory requires 80 of the microsubsystems previously described mounted on two cards.
It provides access time in the order of 100 nanoseconds.
$13,000 is included for disc drive and one on-line disc
pack. The interface with a human operator is through
a keyboard costing $100. A medium-speed line printer
costing $400 and a display assumed to cost $250 are
assumed. The proposed display will make high usage
of special microsubsystems to achieve low cost. The
total cost for this system is estimated to be $23,450.
HARDWARE CONCLUSIONS
The conclusions one may draw relative to the hardware portion of such a system are:
• Hardware costs insignificant ($23,450 is less than
the salary of one professional by 1970).
• Need high volume.
• Need effective problem solving tools (language).
• Need new software (systems).
• Need better man-machine communication (equipment).

CPU (~7094)
FAST STORAGE
BULK STORAGE
KEYBOARD
PRINTER
DISPLAY

1,200.

a,500.} 21,500
13,000.
tOO.}

400.
250.

750

TOTAL $23,450.
Figure 4.

an alphanumeric keyboard and a hard-copy line printer.
Also included is a graphic display which is capable of
producing analog as well as alphanumeric displays.
This system can do both commercial and scientific
problems ranging from small to medium large. This
local data processor will also tie into a telecommunications network so library routines and data may be
swapped. We envision the system doing its own computation for all applications. except for a few types of
large problems. This hypothetical system is not proposed to be used for bulk file processing such as social
security data processing, large insurance company files,
accounting information for large businesses, etc. It is
presumed large file processing will be handled in a
batched manner in' large computational centers.
The cost of such a system is made up of the following
elements: $1,200 is for central process unit logic;
$8,500 to provide 4,000 words of very-high-speed

First, $23,450 is apt to be less than the salary of one
professional person by 1970, even excluding overhead.
The computation facility such a system will provide,
will make the hardware cost insignificant. To a first
order approximation, one may assume hardware costs
are zero and then concentrate on the use of such a
system and how to obtain software. To achieve this
cost, this type of system must be in high-volume production. In order to reach such a production, two
items must become paramount. First, this system must
give the problem originator effective computation aids,
in his own natural language and at his desk, in order
to boost the demand. Second, we must develop an
adequate software-hardware combination to provide
real computational assistance. If this means lavish use
of low-cost hardware to make software development
simple, then we should certainly start thinking in this
frame of reference. Having postulated this technically
achievable objective by 1970, one may ask these questions, "What will it do to help the problem originator?
How can we possibly develop the programming necessary to make it run?"

THE MAN-MACHINE INTERFACE
Keith Uncapher
The purpose of this panel is to encourage creative
use of $.02 logic elements, to stimulate rapid growth
of computer technology, and to provide low-cost computational assistance to anyone who can use it.
This paper discusses assistance to a particular class

Promising avenues for computer research
of computer users-namely, the practicing 1,000,000
engineers in the United States. Though not necessarily
the most important class of users, it is one which many
of us know a great deal about. Throughout this paper,
"the user" or "the casual user" generally means the
noncomputer-specialist practicing engineer. One promising avenue of research is to consider how to give engineers (and others) low-cost, on-line, on-demand,
computational assistance.
First, let's examine the practicing engineer's viewpoint of the computer equipment and systems provided
for him during the last 16 years-that is, the viewpoint of the casual user, the noncomputer-specialist,
nonprogrammer engineer interested only in the tools
available to help him solve his problems.
In 1949, the engineer could be on-line to an analog
computer: he could sit at a console and interact with
the machine. The graphic input/output, dials, meters,
and gadgets helped him formulate and solve his problem. More important, the engineer problem-solver at
the console of an analog computer was part of a very
tight feedback loop including himself, his problem, and
the machine.
These were all advantages for the engineer, but the
situation had negative aspects too. The user saw too
much unfamiliar hardware, which often impeded the
solution of his problem. He could easily touch + 100
volts while altering wires on a plugboard. Further, he
had to know the layout of plugboards and the nature of
the instruments, bells, and whistles which surrounded
him. To be an effective user, he even had to understand
the guts of the analog computer-drift rates, ground
loops, how an operational amplifier could integrate, etc.
-because a direct relationship often existed between
the problem being solved and the stability of the overall analog system. That is, the stability of the computer
could greatly alter the accuracy of the solution.
Thus, the user was forced to pay close attention to
the system itself rather than to just solving his problem. Figure 5 depicts, in a gross sense, the user's view

ANALOG COMPUTEIL

Figure 5.

of an analog computer. Note that while he faced some
"impedances" (i.e., too much hardware, system stability problems, etc.), he was in fact totally and continually
coupled into the feedback loop including the machine
and his problem. Graphic aids allowed him to continuously monitor the behavior of his problem. But in
spite of all the positive advantages of a well-designed
analog computer, few engineers used them. Many engineers in the late forties apparently were unwilling to
learn enough about analog computers and computing
to solve their problems.
In the early fifties, the advent of 701's and 1103's
tended to push the analog computer into the background. Many installations attempted to give each user
a turn at the console of a 701. For the skilled programmer, the 701 was indeed a blessing-but not for the
casual user interested only in solving a problem. Fig. 6

MACHINE LAN6UA6E
OPE~ATIN6

IN
MACHINE ROOM

PO£SSIJ~E Of 0PEDAT1N6
AT CONSOLE OF MACWINE

ENVI~ONMENT
~EADOOT

CONSOLE D£5I6HED fOG
SOE(IAUSTS

NOT

COMM£~

/liT USEIYS SID£

...-I

MACHINE
LANGUAGE

U5fQ'S QUEUE LI/(ELV
I

I
--...I

Figure 6.

depicts his probable reaction. Such an operation required the user to deal with machine language and a
very high-priced machine; therefore, time at the console was precious and costly, creating pressure on the
user. Moreover, he often had to wait in a long queue.
To effectively use the machine, he again had to know
something about the hardware with which he was dealing. The fact that the word length was 36 bits, the
read-around ratio, the expected reliability of tapes, etc.,
were factors which determined the kinds of problems
that the system could solve effectively and reliably.
Again, much of this distracted from the major reason
for being at the console-namely, the solution of a
problem. Operating at the console of a 701 was extremely attractive to the computer specialist or to the
skilled programmer, but not to the casual user.
Although many engineers shunned early digital computers, those who did use them clamored for more and
more console time and computing. The need to serve
more users stimulated the development of batch pro~
essing.
Batch processing brought higher-level machine languages which both benefited and hindered the casual

Promising avenues for computer research

user. Again, the long queue is present, along with new
"impedances," such as long turn-around time and
occasionally dealing with a nasty operator. The user
needed to understand less of the machine's internal
hardware but, to be effective, still had to know the word
length and many other machine parameters. He had
to memorize books of special rules for higher-level
languages. Again, for the casual user (see Fig. 7), we

BATCU PIlOCESSING-DIGITAI" COM PUTEIl

-0r"I.~

USE~

DEALING

LONG QUEUE

WITU OPEN

LOOPSVSTEM
SYSTEM OFTEN
DEMOTE FOOM
USE~ OffiCE

The rest of this talk defines promising avenues of
research which, hopefully, will attract those currently
unwilling to use computers. One such avenue is to design systems tailored to the needs of the casual user
rather than computer professionals. One of the greatest
potential markets for "free" hardware is low-cost,
casual-user-oriented systems which can solve a variety
of problems for many classes of users. Present computer technology is probably sufficiently mature to be
able to build a few such systems. In considering their
design, keep in mind that casual-user-oriented systems
should:

LONG TU~ AQOUNO TIME
COMPUCATED PlWCESS

HIGH LEVEL
TO GET .lOS IN SVSTEM
LANGUAGE
usm MUST BECOME
FO~ THE
....._ _-1 P~OFESSIONAL ...._ _--!! ~ OF SYSTEM

1..- __ -

s~~~~~

___ J;

Figure 7.

surely didn't do very well.
During the past three years, the rediscovery of being
on-line has given new hope to the casual user. Once
again he can work from a console, as in the former
analog and 701 days. Although the emerging on-line,
time-shared systems are exciting, promising, and productive, few cater to the needs of the casual user.
Many are retrofits of languages designed for relatively
skilled programmers. Most do buffer the user from
much of the hardware and some of the software of
the machine with which he is dealing, but many still
require him to know a great deal about both hardware
and software (Fig. 8). For the computer specialist and

PRESENT ON-LINE TIME SllARED SYSTEMS

LANGUAGES PDIM~ILV
FOR PROFESSIONAl..

PROG~MED

\V&TEM"QUL£S

COMPLICATED

CONSOlE

SVSTEM NOT DfSI6NEO
FOil CASUAL usm

1. Allow bilateral conversation in a language of the
user's choice.
2. Require little or no training.
3. Provide the user with adequate and constant
feedback regarding the formulation and solution of his problem.
4. Be more helpful to the user than such alternatives as FORTRAN, slide rule, desk calculator,
service programmer, computing aide, etc.
5. Not be fragile or inflexible in the hands of the
user but, rather, resilient and responsive.
6. Allow the user to interact via a "personal"
console in his office, his home, or both-and
probably with graphic capability.
7. Must allow user-system interaction data to be
stored on-line with nearly instant access.
8. Assist the user in the formulation as well as
the solution of his problem, and permit him to
concentrate on that solution rather than system
mechanics.
9. Cost so little that long periods of console inactivity are acceptable.
10. As much as possible, "seal off" the user from
the messy details of the computer hardware and
software.
In a total sense, the user should find the system a
"helpful assistant." Figure 9 depicts a user's view of

CASUALwUSER oalEImD ON-UNf SmEM
.,~

SOFTWAUE

Figure 8.

the skilled programmer, it's no problem. For the casual
user, it is a level which many will never bother to attain.
Just being on-line is not enough to attract most casual
users.

Figure 9.

Promising avenues for computer research
such a system. It shows that, although some of the
"impedances" of other systems are not eradicable, at
least the messy technical aspects of the software and
hardware should and can be masked: The user should
see only what helps him solve his problem.
Let's consider other promising areas of research.
Certainly it is important for the casual user to be online in the familiar surroundings of his own office. All
of us have experienced the trauma of trying to learn to
use a computer in the presence of skilled, professional
programmers. The novice is more at ease-and thus
more likely to not give up-in the privacy of his office,
learning via a "friendly" console, without an expert
looking over his shoulder. Therefore we must provide
the user a "personal," low-cost, reliable console in his
office or home. It must cost so little that it becomes
another piece of office furniture-not removed from
his office if he goes on vacation. The user should be
able to justify the console and his use of the system
on the basis of a few hours of use per week.
A typewriter is initially an adequate input device
for many users. Eventually, sophisticated languages
will allow the casual user to solve more difficult problems, creating a demand for graphical input/output
devices and graphically oriented languages. To be helpful, the level of interaction through graphics must be
so intimate that many users will need a sophisticated,
high-speed (expensive by today's standards) graphic
capability. Another promising avenue of research is to
build low-cost, highly interactive displays-say less
than $20,000 per console. Perhaps the broad application of digital techniques in the design of displays could
lower the cost and improve performance.
Certainly "free" hardware will permit significant
processor capability on the back of a typewriter or
display. The equivalent of a CDC 3200 on the back of
a typewriter will be possible by 1974. If research provides what the casual user really needs, he will probably not be satisfied with just the ability to communicate
directly with a modest computer. The engineer at the
console will probably want to communicate with his
friends at other consoles, with data bases, and with
other larger and faster machines. Thus, low-cost, cheap,
reliable communications are necessary to connect stations, to transmit anywhere quickly and cheaply-in
fact, it is required to further the use of existing on-line
systems. To put it another way, present on-line system
use is severely hampered by the lack of low-cost suitable
communications. Present common carriers employ slow
switching techniques. Moreover, if the user wants to
buy some time, he must buy at least three minutes'
worth, which is enough to transmit 126,000 bits (over

a slow-data-rate line), though he may want to pay for
and transmit only 3 bits. We can do better.
Skilled digital computer specialists consider a promising area of research to be the design of an all-digital,
nationwide communication system required to connect
users, consoles, computers, and data bases in any way
required-cheaply and quickly. The concept is sensible,
and such a network is a prerequisite to full utilization
of on-line, time-shared computers.
In summary, there is plenty to get excited about.
Every engineer, programmer, businessman and student
can be on-line. However, we should move slowly until
we are sure we know how to design and implement systems for the casual user. We can learn his needs and
requirements, but not easily. Once we do, the technical
progress of the past 15 years will permit us to design
and build new systems far surpassing our accomplishments to date.

PROGRAMMING RESEARCH
Tom Steel
The promise of essentially "free" hardware presents
an important challenge to those who produce software.
The point has been made that the cost of hardware will
become insignificant in terms of the total systems cost.
The cost of programming, by today's standards, is
enormous in contrast to "$ .02 logic." In fact, it already
appears to be true of many big new systems that the
cost of preparing the software perhaps equals that of
the hardware (at today's relatively expensive hardware
prices). If we are to make this "helpful assistant" work,
it will be necessary to get repetitious software costs
down to some reasonable figure. It is therefore important that we of the programming community forget
many of the constraints that we have placed on ourselves for one reason or another. For example, in
today's batch processing environment many people
spend a great deal of time and effort attempting to
make their programs "short" in terms of the storage
space they require. All they need to do is make them
short enough to fit into the high-speed memory of the
machine.
One of the reasons that computer systems do not
now have some of the features that a "helpful assistant"
should have is that (in terms of our present view of
the way machines operate) these features cost too much
in machine time to run. But suppose machine time itself were negligible in cost. We could then parallel
operations sufficiently to get real time down to a
reasonable point, and so, if overhead didn't cost anything, who would care if it were 20%, or 50%, or
95% of total running time-if it served a useful pur-

Promising avenues for computer research

pose. I think a reasonable guess might be that it
would be necessary to spend about 50% of the machine's time and capacity on overhead functions. That
may sound like a lot, but the situation today is almost
that bad in many cases.
What can be done to provide the forgiving, friendly,
interactive, "helpful assistant?" What are the sorts of
things we need? One thing we don't need is the FORTRAN compiler or the PL/I compiler. We must have
some much simpler, problem-oriented language-more
like the kinds of languages, notations, etc., that the user
ordinarily employs. If these users are to be "a million
engineers," this language looks something like mathematics, but you will also find a great deal about format,
loops, and other things that the casual user doesn't understand, won't bother to find out about, and shouldn't
be asked to. It is necessary to eliminate this kind of
thing, and it is important to recognize that much of this
elimination can be done today. We haven't don'e so yet,
and there are some good reasons-mainly related to
cost. We haven't attempted to integrate the various techniques that are available for solving problems. We do
not have systems that have a simple, quick calculation
mechanism coupled with a data file retrieval system. We
do not have a formula manipulation program together
with something like an analytic differentiator where
the interfaces are compatible. But there is no really
good reason why, with careful design and thinking
ahead, it isn't possible to put these things together into
a comprehensive package.
We would need a set of basic mathematical functions, some sort of generalized array processor, facilities
for the generation of subroutines that would compute
special functions, and algorithms for going from a series
definition of a function to a rational function which
could have the coefficients optimized. Further, it's
perfectly feasible to go from the last-named item to
the actual generation of the machine code to compute
the function given the arguments.

of the things that engineers would like to have available
to them is some kind of reasonable fact retrieval system
so that they don't have to rush off to the library to
get the handbook they didn't happen to have on their
desks. This ought to be available through the console.
One should be able to approach this retrieval system
through the use of natural language questions; so the
system should provide, additionally, the facility for
natural language inquiry. Again, we know pretty much
how to do this today. There are some ambiguity problems in handling natural language inquiry, but most of
the ambiguity problems can be easily resolved provided
a dialogue exists. If there is a possible ambiguity, the
system comes back and says, "Did you mean this or
that?" and it doesn't do it in a very demanding way.
Undemandingness is part of this matter of being "forgiving." If you try to intimidate the user by harsh,
cryptic, or telegraphic messages, he is not going to be
very happy about it. He may not know why this
bothers him, but it will, as a good deal of evidence
shows.
What does all this take in terms of today's programming? A rough guess is about 500,000 instructions of
the sort that are in the machines we use today. That is
a lot of programs. I looked at the question of how much
effort it would be to design, implement, and check out
such a system in terms of the programming alone. It
is a job that could quite likely be done in about 125
man-years of effort and could probably be done right
if you gave it five years and didn't try to meet some
irrelevant deadline. That may seem like a huge effort,
but it is quite clearly less than the effort that IBM is
currently putting into the production of 360 software,
and it would be much more useful.

It is then. necessary to have all of this automatic.
You can't ask that the whole proceeding come to a halt
while the user puts in a few cards that perform links
between one set of programs and another. The interfaces between the subsystems have to be designed
smoothly.

So far I have discussed what we need and how much
we have at present. What is required in the way of research? Most of the things I have discussed already
exist; it is a matter of putting them together. It is at
these interfaces between the programming subsystems
that we need the research. Also, we need a great deal
of study at the man-machine interface. We need to
human engineer the programs-not just the console.
It may be important whether a console display is angled
at 30° or 60°, but it is also important how many
characters there are in a message and how many different characters are available.

In addition to this, one needs some kind of reasonable display and printing program. One should be able
to describe formats in a very simple, probably pictorial
way. Then, of course, one needs some kind of monitoring system on top of all this to tie it together. One

There is much work here that needs to be done, and
so far it is not getting adequate attention, although I
begin to detect signs that people are recognizing this
need and beginning to think about studying it. Given
five years, we ought to have most of the answers.

Promising avenues for computer research
POSTULATED SYSTEMS
L. C. Hobbs
The title of this portion of the panel is somewhat
misleading since I will be discussing postulated improvements in systems rather than postulated systems.
Emphasis will be placed in promising avenues of research and development from the systems standpoint.
Two promising avenues that offer significant system
improvements-programming research and man-machine interaction-have already been covered by Tom
Steel and Keith Uncapher. The low-cost storage and
logic elements resulting from batch fabrication described by Rex Rice are certainly essential to both of
these. Low-cost batch-fabricated memories will make it
feasible to store the large program libraries called for
by Tom Steel's postulated programming system. Small
low-cost local stores will make it possible to store
frequently used programs or program segments locally,
thus requiring relatively infrequent access to a large
centralized program storage. In this regard I think
programmers must be very careful to assure that new
software for systems implemented with new technologies is not adversely influenced by 'past experience
with old technologies.
To a large extent these low-cost batch-fabricated
elements will also make possible the low-cost displays
necessary to effect the close man-machine interaction
called for by Keith Uncapher. At least the storage and
logical functions of the man-machine interaction console can benefit significantly. If we do not place too
stringent requirements on the visual image generation
portion of the console and if input techniques such
as the RAND Tablet are effectively utilized, a low
cost display console should be feasible. After all, a
television set that utilizes no batch-fabricated electronics and that includes RF, IF, and audio circuits can
now be purchased for under $100. In more sophisticated man-machine consoles, communications costs
will probably force inclusion of a general-purpose
stored-program computer in the console.
First, I would like to discuss very briefly some areas
of research and development related to computers and
central processors that offer promise in the short range
and then some much more significant areas for the long
range in which the computer and central processor are
negligible factors. It should be emphasized that avenues
of research and development discussed here are promising in the sense of potential payoff if the problems can
be successfully solved, rather than promising from the
standpoint of likelihood of success.

SHORT-RANGE PROMISE
Utilization of Large Arrays from Fabrication
and Interconnection Standpoints
Utilization of very large functional arrays of interconnected circuits is essential to achieving the very low
cost potentials discussed by Rex Rice. This is also
essential from the standpoint of improved reliability
and maintainability. However, one is faced with two
major problems in considering large arrays:
The possible need for eliminating bad or substandard circuits from the array to achieve a reasonable yield.
The lack of flexibility resulting from large arrays
which tends to make each array within a system
unique.
There are three major approaches to the utilization of
large interconnected arrays from the systems standpoint
that are under consideration. The first is cellular logic
in which large arrays of identical circuits are fabricated
with a standard interconnection pattern (e.g . , connecting each circuit only to its four adjacent neighbors)
with the ability to modify the function of the circuit by
changing something in the ·~ircuit subsequent to fabrica..
tion. For example, one approach of this type uses a
circuit with four cut-points which can be cut in differ..
ent combinations to alter the function of the circuit.
In the second approach, a large array of circuits
is fabricated and each circuit is individually tested. The
test results are put in a computer which is also storing
the logical equations of the function to be implemented.
The computer then generates the proper interconnection
pattern to interconnect available good elements (skipping the bad ones) to perform the required logical
function. In this approach, a separate mask must. be
prepared for each array fabricated; hence,this is an
expensive operation unless cheap methods can be developed for producing interconnection masks under
computer control. Several such mask fabrication techniques are under development. On the other hand this
approach offers a major advantage since it is easy to
vary the function performed by the array by changing
the logical equations supplied to the computer. If each
interconnection mask for each array is generated individually, there is little incentive for rigidly standardized functions.
The third approach is advocated by those who believe that in the future it will be technically feasible to
achieve high yields of large integrated circuit arrays in
which all circuits are good. This would permit a standardized interconnect pattern to be used for each specific

Promising avenues for computer research

logical function. This has the advantage that only one
mask need be made for a particular function. This mask
can be used to interconnect the circuits in many arrays
of that type. On the other hand, it is more difficult to
change the function to be performed by the interconnected circuit array since this requires making a different mask.
Since I believe that the integrated circuit industry
will achieve success in either the second or third approach, the cellular logic approach appears to be of
only interim interest as a solution to the problem of
achieving a reasonable yield in large arrays of interconnected circuits. If cellular logic approaches to machine organization find a place in the computer industry, it will be for other reasons. When the second
and third approaches prove feasible, we will probably
find fixed interconnect patterns used for arrays produced in relatively large quantities where flexibility is
not as important-i.e. highly repetitive functions such
as storage arrays, registers and adders, etc. Computercontrolled variable interconnect techniques will be used
for low-volume applications such as the control portions
of a computer where each function tends to be unique.

Highly Functional Organizations
In order to facilitate the use of large arrays and to
minimize the interconnections between batch-fabricated
units, extensive research and development in new machine organization and system design techniques is
needed. Logical design and machine organization approaches must be developed that will permit a machine
to be organized along much more highly functional
lines than at present so that a computer can be assembled from a limited number of relatively large functional blocks. These functional blocks should be selfcontained to the greatest possible extent with a minimum number of signals passed from one functional
block to another.
With the low-cost elements postulated by Rex Rice,
such functional organizations can be achieved at the
expense of quite inefficient logical design within each
block. The number of logical elements within each
functional block will be relatively unimportant. The
number of interconnections between functional blocks
will be of major importance. Hence, logical designers
need to concentrate on minimizing interconnections between batch-fabricated arrays rather than on minimizing
flip-flops and gates as in the past. The criteria for
machine organization and system design should be
adaptability to batch-fabrication technologies rather
than logical efficiency or minimization of logical components.

Modular Computers and Multi-computer Systems
The development of modular computers and multicomputer systems offers other approaches to the standardization of relatively large modules within the system.
If the machine is organized along highly functional lines
as discussed previously in order to use very large arrays,
there will be a strong tendency for each functional
block to be unique in a single computer.
Two different approaches may offer promise in this
area. One is the design of highly modular computers in
which a given module that performs some specific
function is relatively slow but is repeated many times
in the system to provide a high over-all systems speed.
The Holland and Solomon machines are examples of
this concept although I am not suggesting those specific
designs as the proper solution to the need for highly
modular organizations from this standpoint.
Another approach, multicomputer systems, has been
under active investigation for a number of years and
several successful systems have been designed on this
basis. The Navy Tactical Data System, which provides
for up to three computers working together on a common problem with direct data interchange, is a primary
example of this type system. However, multicomputer
systems to date have been limited to a relatively small
number of computers in each system. To aid in the
problem of utilization of la:r;ge quantities of identical
batch-fabricated arrays, it will be necessary to think in
terms of multicomputer systems utilizing a large number of very small standardized modular computers. This
not only raises machine design problems but also severe
programming problems requiring software research and
development to permit the effective utilization of large
numbers of small standard modular computers in a
mUlticomputer system without prohibitive overhead
costs for executive control routines.
LONG-RANGE PROMISE
The avenues for research and development discussed
above are considered short range since it is believed
that reasonable solutions will be found within the next
few years. From the long-range standpoint, these areas
will not present significant problems. If we accept Rex
Rice's postulate of essentially zero cost logic and storage elements for the future, it is a reasonable next step
to postulate essentially zero cost computers and internal
storage.
This brings us face to face with the hard fact that
reducing the costs of the central processor and internal
memory to zero would probably reduce the total system
cost by less than 50%. This ratio will vary of course

Promising avenues for computer research

for different type systems, but looking at the broad
range of all types of information processing systems
40 to 50% is probably a good average estimate. Reducing costs of the low-level digital logic and storage
portions of peripheral equipments to zero might result
in another 10% reduction in systems cost. We are still
left with roughly half the systems cost that is not affected by the advances in electronic and magnetic
batch-fabrication technologies and the zero cost elements that Rex Rice postulated. Hence, from the longrange standpoint, this half of the system presents the
promising areas of research and development with respect to the potential payoff if successful. I have no
real solutions to offer but would like to outline three
major areas of this type and suggest some possible approaches.
Input/ Output Equipment
The most significant problems in future systems will
be encountered in the input/output area. There are
three major approaches to improving the performance
of future systems with respect to input! output equipment. These are:
Improvements in the performance of present types
of input! output equipment.
Development of new types of input/output equip.~ ment that are not in widespread use at present.
System organization approaches that minimize the
need for conventional input/output equipment.
Each of these approaches will play a part in performance improvements in future systems. However,
unless much greater effort is placed upon the development of nonmechanical input! output equipment, the
best hope for future systems probably lies in developing
system techniques that minimize the need for input/
output equipment.
Improvements in Conventional Types of Input/Output Equipment. Almost all present types of input! output
equipment involve electromechanical techniques and
components to a large extent. Many also involve highvoltage or high-powered electronics which are not
amenable to batch-fabrication technologies. This imposes limitations on the improvements that can be
achieved and on the ability to utilize the benefits of
batch-fabrication of electronic and magnetic components. Although these electromechanical input/output
equipments will limit systems performance, the effect
on systems cost and reliability is even more serious.
The performance limitations could be overcome to some
extent by using a larger number of input/output units,
but this further accentuates the cost and reliability im:'"

balance with respect to the central processor and
memory. Performance improvements of less than one
order of magnitude, and in most cases of less than
two-to-one, over equipment commercially available today are anticipated.
New Types of Input/Output Equipment. Several new
types of input/output equipment are under development that offer promise for performance improvements
in future systems. These include:
Character recognition and print readers
Voice recognition and voice output
Nonmechanical keyboards
Graphic input and output
Solid-state replacements for magnetic tape equipment
Some of these, such as optical character readers, are
in limited use at present, while others, such as voice
recognition equipment, are probably 10 years or more
away. Advances in integrated circuit logic components
and batch-fabricated memories will provide significant
reductions in cost of flexible character recognition and
voice recognition since the implementation of equipment of this type involves complex logical functions.
The important role that keyboards have always played
as a man to machine-language transducer will be facilitated by th~ development of new types of keyboards
that do not involve mechanically moving parts and that
may permit more design freedom from the human factors standpoint. New graphic input devices, such as the
RAND Tablet, coupled with low-cost displays will further facilitate man-machine communication and interaction.
Solid-state replacements for magnetic tape may improve the speed and reliability available for this type of
input/ output function, but cost competition with magnetic tapes is questionable. If solid-state storage modules
that can be plugged into read-write electronics in a
manner somewhat equivalent to placing a reel of tape
on a tape unit prove feasible and economical, the input/
output and off-line storage functions presently provided by magnetic tape could be provided by high-speed
high-reliability devices and media with no moving parts.
A BORAM device of this type, providing random access
to a block of data in the storage module, could also be
used as a replacement for electromechanical on-line
mass memories.
System Organization to Minimize Input/Output. The
greatest improvement in the performance of input/
output equipment can be achieved by avoiding input!
output operation,s wherever possible. By keeping the
data within the system and by capturing data at the

Promising avenues for computer research

source, much of the need for conventional types of
input/ output equipment can be eliminated. For example, the need for voluminous printed reports can be
reduced sharply if the user is operating on-line with
the processor through an efficient console. When any
part of the data base within the system is rapidly available to the user upon request, he will have little need
for large reports that are used for occasionally looking
up printed results-particularly since these may be out
of date by the time they are used. In general, any effort
to increase the extent to which systems are "on-line"
will tend to reduce the amount of conventional input!
output equipment in the system. Achieving the improvements possible in this area will require a combined effort of users, programmers, hardware engineers,
and systems planners and designers.

approach to this. The communications problems takes
on greater importance as we tie together remote computers into a geographically dispersed data processing
network.

Very Large Capacity Mass Memories

In the past, the cost of a large high-speed computer
has been much less per operation than that for a small
computer. As logic and storage elements become very
cheap, this argument falls by the wayside. When it becomes cheaper to provide a stand-alone computer that
permits occasional access to a large centralized data
base than to provide more frequent communications for
a terminal that depends upon a public utility for its
computing capability, we will find small stand-alone
computers in widespread use. If the local drug store or
market can have a competent stand-alone computer
for $10,000, it will not be economic for them to tie up
a telephone line for continuous communication with a
centralized computing utility unless significant improvements are made in communications techniques.

Very large capacity mass memories ~epresent anmajor problem area for future systems, which is
closely related to the input/output problem discussed
previously, since similar techniques and mechanisms
are used at present. Batch-fabricated electronic and
magnetic technologies will provide solid-state on-line
auxiliary storage with reasonably large capacities in the
order of 108 bits., Cryogenic techniques may push this
up to 109 to 1010 bits. However, for very large auxiliary
storage requirements in eXcess of 109 or 1010 bits,
electromechanical devices will be required for the foreseeable future to keep the costs within reason. Of
course, this somewhat contradicts the points made previously under the concept of zero costs storage· elements,
but almost zero cost per bit multiplied by 109 bits may .
still be an undesirably large dollar figure. Even if solidstate random-access on-line auxiliary storage costs drop
to something in the order of 0.1 ¢ per bit, electromechanical auxiliary storage will still be required to
achieve costs in the order of .001 to .01¢ per bit for
very large capacity storage. Further research and development in block-oriented random-access memories
(BORAM) may eventually provide a solution to this
problem that will eliminate the necessity for electromechanical mechanisms with consequent improvements
in cost, size, and reliability.
o~her

Communications
As the costs of digital elements and equipment continue to drop significantly, we will quickly reach a
point where communications costs become a major
problem unless improved communications techniques
are developed. Keith Uncapher has pointed out one

Those proposing "computer utilities" appear to sweep
this problem under the rug a little too easily. If you
give the average housewife a free terminal, she probably cannot afford the telephone costs to tie into the
computer utility in Chicago or Kansas City. The only
really incontestable argument for the computer utility
concept is the need for a common data base (where
the "data base" includes both data and programs). All
other arguments for a computer utility are really based
on economic decisions that will vary as technology
changes.

fortunately, the concept of zero cost logic and storage elements offers some promise for a partial solution
to the communication problem. Keith Uncapher has
already pointed out the effects of more extensive use
of digital communication techniques. In addition, the
relative cost between computing and communication
will change to the point that small users can be provided
their own relatively sophisticated stand-alone computer
and internal memory capable of accessing a remote
large-scale system when necessary. With this approach,
it will not be necessary to provide frequent or high-speed
data communications between the central system and a
typewriter-like terminal for each step in the process.
Instead, the user's own stand-alone computer will perform most of his work. Access to the central large-scale
system will be requir~d only for certain data or programs and some types of calculations that may be too
complex, for the small computer. Computing and storage capability in' the terminal will minimize the data
exchange required with the central system.

Promising avenues for computer research
Conclusion

In closing, I would like to summarize some specific
avenues of research and development from the systems
standpoint that appear promising with respect to potential payoff.
• Low-cost BORAM or other types of large capacity
. alterable on-line storage.
• Low-cost and more useful display consoles compatible with integrated circuit and batch-fabrication technologies.
• Low-cost "true" input/output (e.g., printers, keyboards, graphic input, optical character readers,
voice input, etc.).
• System design and application concepts that minimize input/output operations at the expense of
more internal logic and memory.
• Modular computers and multicomputer systems.
• Improved concepts of increased functional modularity.
• Machine languages that facilitate compiling operations.
• Improved automatic fault isolation and maintenance.
• Techniques for minimizing communications requirements.
I am sure that each of you can compile a longer list
in your respective areas. The promising avenues of
research and development from the standpoint of potential payoff are those directed toward the problems
that will limit the capabilities of future systems. Unfortunately, these are the difficult problems rather than
the glamorous ones.

PART II
DISCUSSION
Note: The space available has limited the published
discussion by the panel as well as by the audience. The taped audienc·e discussion has been
condensed and edited. The panel has tried to
include the principle thoughts of the audience
but has had to reduce verbage. We apologize
in advance if we have unintentionally misinterpreted or misrepresented the remarks by
members of the audience.
IVAN FLORIS (Stevens Institute): . . . I don't think
we can solve a universal language for engineers and
scientists. I think that we should ask more of them to
try to structure their problem in a language which is

not a natural language but rather a formal one. If we
try to make it too easy for them, we are going to find
out that we have intellectual machines dealing with
more professionals and I don't like that idea.
TOM STEEL: I think perhaps the best response to that
business about getting trained professionals to all use
a formal language is that (I agree) it isn't hard for us
to learn FORTRAN, it is just a damn nuisance. It
probably isn't hard to learn Swahili either but I don't
want to bother. It is all well and good to talk about
requiring that the professional learn precision, learn
how to state his problem properly and to be formal and
precise when it is necessary, but there is such a thing
as being formal just for the sake of being formal and
the kind of languages that we have now put an excessive burden on the user to take care of grubby details associated with solving his problem that he ought
not to have to bother with. That is, one of the reasons
these machines were designed and built was to relieve
the man who was solving a problem from fussing around
with all the grubby details of adding 2 and 2 and getting
4. Now we'll just carry it one step further. There's a
lot of mechanical organizational elements in sequencing
a set of calculations that the user shouldn't have to fuss
with. He has better things to think about.
CARL BROOKS (Northrop Corporation): I would like
to make a comment based on that-a question which
I think perhaps is addressed most directly to Keith
Uncapher. In trying to build tl!e locked box (Fig. 9),
for which there is no need for the user to glimpse inside,
I don't disagree with this as a goal-I think this is a
fine goal but I would like to draw one of the many
parallels that one can make with current analog and
digital machines and from there sort of point out a
paradox with this situation. The range of one common
breed of analog machine such as Keith mentioned is
from -100 volts to + 100 volts. One has the parallel
situation in a digital machine of a given six-bit word
length. Scaling is always a problem and no matter what
engineer you are trying to turn this problem over to, he
can always find a problem that is going to blow your
scaling out no matter how long you make your word
or how many volts you make your range on the analog
machine. Hence, when you get to that problem (and
we in the research business are faced currently with
that problem as much on a digital machine as on an
analog machine), you come to the point where you
have to look inside the box. You have to find out
what are your eventual hardware limitations. Nobody
has built a software package yet that will overcome
this. Now I would like to ask Keith what prospects he

Promising avenues for computer research

sees for the development of a new outlook in software
or a new outlook in hardware-a complete breakthrough into a new kind, an entirely new concept in
computing that would overcome these problems which
run parallel in both analog and digital machines.
KEITH UPCAPHER: I am not sure that I see any
breakthrough, but what I do see though is the ability
for us to understand the needs, let's say, for the engineer and to provide for him the appropriate feedback
and limited formatting of information. The total system
design and philosophy should be such that the user is
alerted if he attempts to exceed the bounds of the system. But within the bounds, he should constantly see
precisely 9-place accuracy, just as if he were going to
an engineers handbook. This is really difficult to come
by at present. That is, it's difficult to design a system
which takes care of the messy round-off problems, yet
seals off the user from the messy details of the computer. I think Cliff Shaw did a beautiful job with JOSS
in this regard. Cliff arranged the language so that
there is no payoff for it and really there is no way to
look into the box and decide if it is a 36-bit machine,
whether it is binary, or whether it is anything.
JEAN SAMMET (IBM): My comment and implied
question is directed to Tom Steel. While I agree with
almost everything that he said, I have to very seriously
question the numbers that he used because I think they
are all off an order of magnitude and unfortunately
in the wrong way. I am perfectly willing to grant some
deviations, but Tom used the figure of approximately
500,000 instructions, today's instructions, for the package that he was talking about. I will point out that I
believe the amount of sfotware that is in the current
project MAC system, not the users' program, but those
things which are in some reasonable sense considered
part of the system itself comprise approximately 500,000
instructions right now. I think that just about everything in there will be needed if this integrated system
and probably at least as much more. I think it would
probably be not too unreasonable to say that you are
off by a factor of about 10. I would be much more
inclined to believe that the number of instructions would
be in the neighborhood of five million rather than half
a million.
REX RICE: During the panel session, I think Tom
Steel brought up a point concerning the environment
that was required as well as the requirement for the
program.
TOM STEEL: . . . I think the direct answer to the
implied question you ask is, "I need a check for 10

million dollars." I think you need about 20 people,
programmers, and need to give them 5 years and detach them from having to do 97 other things so that they
can concentrate on doing this job. You can't totally
isolate them because in order to do this right, they are
going to have to have access to guinea pig engineers to
tryout some of their ideas. I think a very important
aspect of this is that the people who design the programs also write and check out the programs. I have
seen systems, that have been designed by one set of
people, coded by another set of people. First of all, the
results didn't meet the design specifications and secondly, the resulting program was about five or six
times as big as it needed to be and I think that is where
perhaps we are having some disagreement on the size
of this package. To go straight to the heart of it, I
think MAC is badly done and it could be cut down
considerably.
JOHN REYNOLDS (Stanford University): I can't resist interjecting one remark that 5 million words of
sloppy code may be less expensive than 500,000 words
of well-written code. But more seriously, I am bothered
that there hasn't been a distinction here between languages which are easy to learn and languages which
are easy to use. I think it is perfectly reasonable to
expect a professional engineer to spend as much time
learning to program as to learn to do drafting. . . .
What I am afraid of is that there is going to be an
emphasis on avoiding powerful languages just to make
it simple to keep the languages easy to learn. What is
really important, particularly with the vital class of
problems which are not put on machines, is not that the
people who want to do them are too lazy, but simply
that in present-day languages the coding effort is too
enormous to develop more powerful languages, rather
than languages that are easier to learn. And by power
I want to stress two things: first simple concisenessthe ability to write code rapidly-and secondly, generality. Too many of these problems run across the
rather arbitrary boundary lines that are set up in the
so-called problem-oriented languages so that you may
have to orient a language to every problem differently
unless you set up much wider boundaries.
NED CHAPIN (Consultant): . . . I think it is implicit
in the panel discussion that they must have considered
the tradeoff in hardware and software, but I have
heard them be very quiet on the subject. I wonder if
they could make a little noise.
CHARLIE HOBBS: I think the vote on that is, with
the kinds of hardware costs that we are looking at for'
the future, the tradeoff comes in favor of the hard-

Promising avenues for computer research

ware. The hardware costs for central process and
internal memory will be so cheap that we can afford
to pay a very high price in terms of number of components to achieve economies elsewhere, such as programming for the user.
BILL HUGEL (Stewart-Warner): Perhaps I have to
defend my own projections, which perhaps in 1970
don't show 3 cents-maybe only a dime-I am not
sure. But in any case, some of the difficulties you face
with this large array having 60,000 devices are due,
I believe, to the fact that you are talking about something maybe of the order of 90% yield or maybe even
99 % yield for a good company like Fairchild. I am
not sure, but if you raise that point nine or point nine
nine to the sixty thousands power, the yield comes
down rather drastically. Then if you take the ten to the
tenth circuits that are required and apply the new yield
factor, I believe that the capacity outstrips even Fairchild and our own put together.
REX RICE: We are not now using redundancy to get
around the yield problem because of the low level of
the circuit sophistication. We are postulating that this
will in fact be required. . . .
CHARLIE HOBBS: There are three approaches that I
am aware of that are being worked on for overcoming
the problem of yields with respect to large arrays. One
is the use of cellular logic . . . in which a large array is
interconnected with essentially a simple and standardized pattern and the price is paid in terms of efficiency
or effectiveness of a circuit. To achieve this, eacl;t circuit has the ability to be modified physically by cutting
or in some way breaking some contacts to change its
function to implement the logical function. The second
one is the approach of using a computer control fabrication of the mask. That is, you test the individual circuits, enter these test results into a computer that also
has a logical equation, generate the program interconnections, then make the mask to connect the good
circuits and leave out the bad ones. The third approach
is for those who are optimistic, and I think there are a
number in the semiconductor industry that feel this
way-that yields will be so good that none of this will
be necessary. . . .
WALTER DOUGHERTY (IBM»: I would just like to
suggest that there may be a point that has been overlooked. I am suggesting that you have a central data
base and perhaps a nationwide central data base or
some large regionwide central data base and a reasonably powerful stand-alone computer which might sometimes access this central data base. I think this is very

reasonable for the engineer and for the problem solver
but I think that one of the major things we face is
handling of information from the point of view of education and from the point of view of simply having
access to information for the drugstore owner, for the
local grocer, etc. I suggest that maybe a further level
should be interjected between the large national data
base and the powerful terminal or powerful local computer. This other level might be sort of a regional system. I am not suggesting a public utility necessarily but
I do think there is a need for a regional-wide data base
and a regional-wide computing facility, where you
would have the possibility of cheap communication.
KEITH UNCAPHER: I don't think I would want to
argue with you. I think either implies a communication
system which has the general properties of those I tried
to describe and something that doesn't exist and I don't
think is going to exist without some fairly prompt action
and a lot of dedication, perhaps by means of us in this
room. The building of a network which will fulfill the
requirements offered by you and others all seem to
suggest that a low cost (per user) nationwide user-user
computer-computer network must exist and I vote for
an all-digital approach.
HAROLD CANTNOR: As one of the many million
potential casual users, I got a bit confused and I wonder
if the panel can orient me in saying whether they plan
to do something for the casual user, or to the casual
user, or with the casual user.
KEITH UNCAPHER: I suggest we do something for
the casual user. . . . I think for the airline reservation
clerk, the industry has provided a fairly good on-line
system which doesn't require the user to know anything
about the system. . . . I hope we can do the same for
the engineer. . . . I think we are beginning to see in
the environment at RAND, with JOSS, that there is a
significant new class of JOSS users who have never
found FORTRAN, etc., worth the challenge. One could
perhaps argue for good or bad reasons that these users
should have used other alternatives, but they haven'teven in a one-hour turn-around time environment.
HAROLD CANTNOR: Do you' belieye that a human
being can truly carryon a dialogue with a machine, or
it is through the technological devices that he carries on
discussions with his fellows?
KEITH UNCAPHER: I think he can carryon a conversation with respect to the solution of a problem
with a computer system-absolutely.
BRUCE McKEEVOR (General Electric Computer Labs) :
. . . The original statement of the panel's objective

100

Promising avenues for computer research

is that they are worried about expanding the market for
computers to handle those 80 to 90% of the potential
users who have stayed away from computers in droves.
Those are just the people that we experts don't have
any, or very little experience with. The airline example
was mentioned. It was possible for the experts to learn
enough about some particular user to develop a system
for him. I am worried about the generality of this being
true and I would like to throw two alternatives out to
the panel to let them chew on it for awhile. One:
Stick fairly close to what you are trying to develop but
instead of developing a system, develop some hardware
and a set of languages for the user, develop some collection of software supplemented by suitable hardware
-that the user can shape into his own system. At the
time he is using it, he can have an inconsistent language
and use it where it doesn't get him in trouble. Where
it does get him in trouble, he can patch his own way
around it on his own terms. Now, an even further out
alternative, perhaps one that is practical gets back to
the question of the free hardware .... I am wondering
if perhaps what we can do is apply this idea indirectly.
... Go ahead and build a batch processing system, a
national one that is huge or a gigantic time sharing thing.
. . . However, let the system be as complicated and
sophisticated, as powerful as is necessary, and require
that the system be exercised by a set of experts, each
of whom now is, in effect, a broker. The customer comes
in, formulates his problem and the user pays the company owning the machine some fee and the expert
(broker), whether he be a programmer or an applica-

tione engineer, gets a percentage of the take....
TOM STEEL: Well, it seems to me that to a not inconsiderable extent that is precisely the situation we are
in now. . . . In the first place, most of these people
for very good reasons don't want to have to be bothered
to sit down and explain what their problem is to somebody else and make him understand it. We have already discussed the fact that he probably has not clearly
formulated his problem and that is part of what he
does when he interacts with the machine directly. It
helps him formulate his problem ... He can do this in a
two-stage process by talking to this paragon of virtue,
the programmer who satisfies everybody, and maybe
he can formulate his problem by interacting with that
guy, but then there is the other step of having to put
it on the machine. It doesn't seem to make much sense
frolJ1 anybody's point of view.
KEITH UNCAPHER: ... First of all, I agree with Tom ..
I think what was proposed is about what we have
today, namely a service programmer of sorts who
also happens to be a broker. If I try to answer for the
engineer for instance, I realize that there are some
exceptions but the engineer, or at least many that I
k now of, looks at the service programmer as a bossy,
bilateral, filter and we would like to get rid of him so
that there could be direct interaction. . . .
REX RICE: On that high ending note, I thank the
audiencc for their very provocative remarks and invite
them to come up and have at us. Thank you.

Late Papers Not
Appearing in Volume 27, Part

A high-speed thin-film memory:
Its design and development
Q. W. SIMKINS

International Business Machines' Corporation Systems Development Division,
Poughkeepsie, New York

Array Design

INTRODUCTION

The memory array is made up of copper bit plates
on which the films are deposited, striplines of Mylar*

Memories in today's high-performance systems are
typically made up of memory modules of capacity
comparable to the new memory to be described here.
, Cycle times are 500 nsec to 1 p.sec with access times
of 300 to 500 nsec. This paper presents the design of
a thin. ::film main memory with a capacity of 8,192
words of 72 bits each. The cycle time of this memory
is 120 nsec with an access time of 60 nsec. Thus, this
memory design represents a 4-to-8-times improvement
in main memory performance over the present state of
the art.

-----,:t=-

,---

Irti
~

~
=-

f - t--

4K
ARRAY

f.._

r--r--

]

-- t--

I--

-~~~~~-

I--

- -=E-

---

::-:

,~~

ARRAY
TO
BOARD
CONN.

l - t--

4K
ARRAY

'--

BIS
WRAPA ROUND _______

~§

II 'rool' 'roo{ 'roo{~
I

68"

SENSE
AMP!:',
REGEN
BOARD

~d~

BIT
DRIVE
CKTS,

The first section of the paper is a general description
of the physical layout of the memory. The memory
array is then discussed with particular emphasis on
noise problems in the array. Finally, the memory circuits and packaging are described.

WORD
TERM,
RES,

42"

I
7"

J:
!

COOLING UNIT

Figure 1.

GENERAL SYSTEM DESCRIPTION
The memory layout is shown in Fig. 1. The approximate dimensions are 68" high by 42" deep by 7"
thick. The memory array consists of two 1,024-word by
288-bit assemblies which are driven and sensed in
parallel. 72 of the 288 bits read in parallel are gated
out. Each assembly contains 72 3" X 3" bit plates arranged in a back to back 6 X 6 configuration. The
dimensions of each assembly are 20" X 20" X 4". The
drive and sense circuits and associated logic are packaged on cards mounted on multilayer boards which
surround the array.

Memory layout-8K gate.

foils, and a ferrite keeper which partially closes the
return flux path of the film memory elements around
the striplines.
The thin-film elements utilized in this memory are
25 X 30 mil rectangles on 30 X 54 mil centers. The
permalloy films are about 800 A thick. They are
switched in ,.., 5 nsec by means of orthogonal drive
fields produced by fast-rising current pulses in two sets
of striplines. The word lines are closest to the film and
are parallel to the easy axis of the film as shown in
*Mylar is a registered trademark of the E. I. du Pont de
Nemours & Company.

103

104

A high-speed thin-fDm memory

Fig. 2. The bit lines are perpendicular to the word lines
and separated from them by a thin dielectric foil. The
sense line is parallel and coplanar with the bit line.
Registration of the strip lines to the bit plates is
achieved by mounting the striplines on rigid frames with
appropriate mechanical registration fixtures. The line
pattern is etched with the copper-Mylar laminate in
tension, and a bit plate-stripline registration tolerance
of + 3 mils is maintained.

rise and fall of the bit current. It determines the recovery time for the array and sense amplifier. The capacitively coupled noise from the bit line divides and is
propagated in opposite directions on the sense line.
Thus, the polarity of the capacitively coupled bit noise
at the two ends of the sense line is the same. The inductive coupling, on the other hand, results in noise of
opposite polarities at the ends of the sense line. Thus,
on one end of a properly terminated sense line inductive
and capacitive noise will add, while on the other end

Array Noise Considerations
The signal energy from a thin-film memory array is
low. In this memory the signal level is 4 millivolts,
signal duration about 5 nsec, and the sense line impedance 50 ohms. At this signal level, array noise
must be minimized if reliable memory operation is to
be achieved. Noise results from unwanted energy coupling into the array at read time and at write time. The
read noise results from capacitive coupling between
the word line and the sense line. Since this noise is
coincident with the signal, it directly affects memory
operating margins. Write noise results primarily from
coupling of energy from the bit line to the sense line.
Since the sense system must recover from the write
noise before the next read operation, the read-write
cycle time is directly affected.
Read noise in an array in which all lines are terminated in their characteristic impedance is given by the
following equation:
CswZowZos
V""

1
PRESSURE PAD

BIT PLATE
WORD LINES

.,,/

MYLAR
INSULATIO N~

I'"

BIT-SENSE
LINES

KEEPER

PRESSURE PAD

Figure 2.

Array sandwich cross section.

Figure 3.

Array plan view-bit and line.

dlw

X {ft

where Csw == the capacitance between word and
sense lines,
Zow == the characteristic impedance of the
word line,
Zos
the characteristic impedance of the
sense line, and
dlw
{ f t - the rate of rise of the word line current.
This noise, unless canceled, is comparable to the signal
level. First order cancellation is obtained by using a
dummy sense line which runs between the bits on the
bit plate. This is illustrated by Fig. 3. The outputs of
the sense and dummy sense lines are sensed differentially using a balanced pulse transformer.
Since the bit and sense lines are closely spaced and
long, the coupling between them gives rise to a large
write noise (30 mv). The reactive component of write
noise results from inductive and capacitive coupling
between the bit and sense lines. This occurs during the

it will subtract. We take advantage of this directional
couplingl by locating the sense amplifier and the bit
driver at opposite ends of the array. The noise reduction' thus gained may be as high as a factor of five. A
second component of the bit noise is due to resistive
coupling in the ground plane. This noise has approximately the same pulse shape as the bit pulse and is
controlled by the magnitude of the ground impedance.
An additional source of write noise is energy coupling
to the sense line from the circulating ground plane currents. This noise is low in amplitude, but since the time
constant for ground plane current spreading may be as

A high-speed thin-film memory

. long as 500 nsec, this noise may be present during the
read portion of the next cycle and so must be considered
in the design of the sensing system.
Write noise may be controlled by a number of techniques. Directional coupling has already been described.
In addition, several balancing techniques are available.
A symmetrical arrangement of bit, sense and dummy
sense lines may be employed in the bit sense lines to
achieve first order cancellation of self-induced bit noise.
This arrangement, however, is wasteful of bit current,
and not effective in eliminating coupling between adjacent bit lines. Both self-induced and adjacent bit line
noise can be canceled by driving and sensing two arrays
in parallel. This arrangement, used in this memory, is
illustrated in Fig. 4.

Figure 4.

Bit-sense system.

Memory Circuits
The word selection system employs a transistor
matrix (one transistor per word line). The nominal
word current is 500 rna with a rise time of 7 nsec.
The matrix transistors and word and gate drivers are
packaged in a card-on-card approach similar to that
employed in Solid Logic Technology (SLT).2 All line
impedances are controlled to provide line matching and
prevent troublesome reflections. The word lines are
terminated in approximately their characteristic impedance (i.e., 17 ohms). Interconnections between the
array and circuit boards are made by means of coaxial
cables.
The bit and sense circuits are packaged between the
two halves on the array. The bit driver circuit provides
100-ma pulses into each of two parallel bit lines with a

lOS

rise time of 7 nsec. These pulses are positive or negative depending on the information to be stored. The
sense amplifier has a matched impedance input fed
from a balanced pulse transformer. The output of the
sense amplifier is gated into a tunnel diode detector,
which feeds a data-register latch circuit. The sense
amplifier, detector, data-register and bit driver are all
packaged on the same board to minimize delay in the
regeneration loop.
The films are vacuum deposited on a 3" X 3" highly
polished copper substrate about 80 mils thick. The
substrate is precoated with a micron or more of silicon
monoxide to smooth the surface and improve the film
properties. This deposition is made in the presence of a
strong, uniform magnetic field and at a precisely controlled substrate temperature to produce the desired
anisotropy in the films. An initial screening of the bit
plates is done by checking magnetic properties using
Kerr magneto-optic measurements prior to photoetching
into the desired bit pattern. Typical properties are
He == 4 oe, Hk == 5 oe, with skew and dispersion less
than 2 degrees.
After etching, the bit plate is pulse tested under
worst-case operating conditions. Each of the 4,128 bits
on the plate is tested with a pulse pattern that corresponds to worst-case conditions, including information
pattern, magnetic history, ground plane currents,
trapped flux effects, disturb conditions, and marginal
currents. Worst-case conditions involve simultaneous
word and bit disturb, including disturb effects from
adjacent lines, as well as those from lines directly over
the bit. A bit plate is accepted only if all bits on the
plate meet the minimum specifications for the bipolar
one and zero signals.
The copper bit plate serves as the ground plane for
the striplines. Interconnection of the bit plate ground
planes is achieved with pressure connections around
the bit plane periphery. To improve reliability of the
ground system, the edges of the bit-plate are rhodium
plated.
A metallic substrate, rather than glass or mica, was
selected for this application for the following reasons:
1. Close proximity of the ground plane to the sense
line reduces noise and permits higher-speed
operation.
2. Lower line impedances reduce power requirements of drive circuits.
3. The ground plane reduces field spreading which
permits closer spacing of the bits.
4. Metal substrate has more uniform temperature
during film deposition reSUlting in more uniform
film properties.

106

A high-speed thin-film memory

Disadvantages of the metal substrate, which include
ground plane current and trapped flux effects, 3 are
minimized by the keeper. The memory design includes
etched discrete bits rather than a continuous magnetic
film for these reasons:
1. Disturb effects ( creep) are less severe with
etched bits, permitting higher packing density.
2. A lower noise-balanced sense system with a
dummy sense line running between the bits can
be used.

Conclusion
This paper has described briefly the design of a
large, high-speed film memory which represents a substantial speed improvement over the current state of
the art in main memories. Key features and design
choices highlighted here include the magnetic film
element design, the metal ground plane substrate, the
use of discrete bits, and the electrical design of the
array to minimize noise effects.
Appendix
FILM MEMORY ARRAY PROPERTIES
Word lines
Zo
170
length
24"
attenuation
<5 %

Bit lines
Zo
length
attenuation
Sense lines
Zo
length
attenuation
Signal
Noise
Read Noise

31 n
46"
<5%

600
46"
25%
4 mv (at the film)
:::: 2.4 mv
0.4 mv
:::: 65 mv
< 30 mv
<

Bit noise

(uncanceled)
(canceled)
(uncanceled)
(canceled)

REFERENCES
1. G. F. BLAND, "Directional Coupling and Its Use
for Memory Noise Reduction, "IBM Journal of
Research and Development, vol. 7, no. 3 (July
1963).
2. J. J. Corning et aI, "Solid Logic Technology: Versatile, High-Performance Microelectronics," ibid,
vol. 8, no. 2 (Apr. 1964).
3. V. T. Shahan, and C. J. Townsend, "Measurement
of Trapped Flux and Ground Plane Current Effects," 1964 Intermag. Conference.

Efficiency and management of a computing system

H. D. HUSKEY
University of California
Berkeley, California
and Universidad Cat6lica de Chile

characters per second. If transferred to magnetic tapes
or disc files, the data rates may be from 100,000 to 1
million characters per second into the computer processor. Automatic operations may be done at close to a
million per second. Then the output process perhaps
involves the transfer of information to a line printer
capable of printing 1,000 to 2,000 120-character lines
per minute. Again, only rarely will all 120 characters
be significant.
From the above discussion it is obvious that throughput, at least for small problems, depends more on
input/output capability than on arithmetic speeds.
In the past, word length in a computer was more related to the number of core planes that could be
driven, and to command length considerations, than to
accuracy requirements in computation. In a university,
most computation is for educational purposes and almost any number of significant figures is satisfactory.
For example, illustrations of loss of accuracy in numerical analysis can be taught as well with four significant
digits as with eight.
The input data for much of the computation are relatively rough, so that computing with eight significant
figures may have no meaning.
Doing multiple precision arithmetic slows most computers by substantial factors. However, if six significant
digits were sufficient for 80% of the problem load
(and. I think it is), then it might be better to put the
incremental cost of more digits of precision into input/
output hardware.
Some computers have been built with inadequate
round-off facilities. Many problems (particularly differential equations) involve millions of arithmetic operations in sequence, so a slight bias in round-off may
accumulate with disastrous effects.

INTRODUCTION
The study of efficiency in computing systems can be
subdivided in the following way: (1) efficiency of hardware, (2) efficiency of software, and (3) efficiency of
operating systems. Efficiency in this context will be
considered as performance for the average customer
divided by potential performance. It is granted that
both "average customer" and "potential performance"
are concepts about which it is difficult to be precise
and rigorous.
Every computing laboratory should use a certain
amount of "instrumentation" so as to be able to define
the average customer and to have some information
about customer distribution. Potential performance depends upon the customer distribution. Care should be
exercised in collecting such information because of
"feedback effects." For example, if it is known that
the use of tapes is being monitored, the users will be
more conservative.
Priority systems enhance the performance for some
customers at the expense of peiformance for others.
Although some aspects of priorities may encourage
more efficient use of the computing system, there will
be no attempt in this presentation to consider their
effect.
HARDWARE
The design of computer hardware systems has been
difficult because of the tremendous differences between
data rates in the parts of the system. Data, in many
cases, originate at manual rates of a few characters
per second. If punched in cards then the cards may be
read at rates of 250 to 1000 per minute. If cards average 30 characters each (a bit high), the rate is 500

107

108

Efficiency and management of a computing system

Almost all computers have index registers which
facilitate the processing of arrays. The real choice here
lies between hardware index registers and index registers in memory. Hardware registers are fast and expensive and complicate the interruption process (since
they must be saved). Memory index registers are cheap,
permitting much larger numbers. They may be incorporated in a paging system which minimizes the cost
of interruption.
Scientific and engineering computation typically involves processing integers and real (floating point) numbers. Some computers have been built so that integer
and real arithmetic are done the same way. This permits
the mixing of types of operands in an arithmetic expression with no extra commands.
In computational processes such as the solution of
systems of linear equations by elimination, the majority
of the commands in the program are involved with the
modification and testing of indices. The inner loops of
such programs ( those commands executed the most
times) can be simpler ( and faster) if flags on the
data can be used to "interrupt" the computing at the
end of a row or column.
SOFTWARE
Most scientific and engineering problems are done
using a problem-oriented language such as FORTRAN.
It has been said that it required 18 man-years of effort
to produce the first FORTRAN translator. At this level
of effort, just any computation laboratory is not about
to modify the FORTRAN translator. Consequently, one
settles for that FORTRAN system which the computer
company produces. Naturally, such a processor has
been designed so as to supply the needs of all customers
in an optimal way. If the majority of the users can
run their problems on a simpler one-pass system (onepass indicates that the source language deck is read
onee, object code generated, and perhaps executed in
place in memory), then the remainder can run using
the full system. Thus, the majority of users are not
paying the overhead represented by the unused portions of the complete system. For example, how many
FORTRAN users use the complex number capability?
Monitors, or operating systems, may not utilize all
the hardware capability of a given computer. This may
be because of size limitations, i.e., the larger and more
sophisticated the monitor, the less space there is left
for the customer in the high speed memory.
Techniques in translating are well developed so that
it is no longer a major task to write a compiler. Consequently, computer laboratories should custom design
their operating systems and perhaps write translators for

languages suitable for the majority of their customers.
OPERATIONAL PROCEDURES
Most computer centers suffer from a lack of information about what is being done by the customers. On
the other hand, it is easy to collect too much information and increase the overhead costs of doing computing.
Batch processing implies a minimum turn-around
usually of some· hours. Unless there are restraints, the
user (1) tries alternative forms of the program-in case
he doesn't understand formats, for example, (2) runs
more cases than necessary just to be sure of spanning
the area of interest, and (3) takes extensive memory
dumps for debugging purposes.
In undergraduate classroom use, one has to choose
between restrictive measures, which minimize extra
runs or extra input and output, or less restrictive measures permitting the student to do essentially whatever
he likes. This last option is not necessarily bad since it
permits the student to explore alternate approaches to
solving the problem.
Magnetic tape is used inefficiently in many computer
laboratories. Many times reels of magnetic tape have
only a few feet with recorded information.
At 500 characters per inch an 80 column card corresponds to about one sixth of an inch of tape. This is
comparable to the start-stop space needed on the tape.
Therefore, if the customer reads cards it is easy for
him to establish 80 character records on tape and lose
50% in efficiency both in terms of quantity and speed.
One may think that disc storage solves these problems. However, disc space may be handled by system
subroutines and these may use buffer sizes which lead
. to similar inefficiencies.
ON-LINE SYSTEMS
The use of consoles on-line to a computer will change
the character of the inefficiencies. Perhaps the most
serious problem is that the user is now encouraged
more than ever to "play around" at the console. With
high traffic usage leading to sign up for time at consoles
this can be controlled some. However, it will be only
user self-discipline in the end that produces efficient
use.
The use of Culler-Freed methods where the user calls
for evaluation of expressions for 100 to 200 values of
an argument will use up computer time at a tremendous
rate. Again the real premium will be on adequate planning so that those minutes at the console will be well
used.

Efficiency and management of a computing system
In on-line systems a great advantage will come from
the fact that the user need not ask for extra information
so the relative printing load will decline markedly.
Since he can sustain attention on his problem he can
complete the task in minutes instead of days and
eliminate reorientation time.
A proper on-line system has to supply bulk storage
for keeping all users' data (programs, numbers, text,
etc.) available on demand. Soon the system becomes

109

loaded with junk. Perhaps user priorities should be
developed which are inversely proportional to the
amount of bulk storage used by the customer.
Except for the use of bulk storage most of the actions
of a console user will have no effect on the efficiency
of use at other consoles. Thus, in contrast to batch
processing, only the individual suffers from lack of
self-discipline.

Use of digital computers in basic mathematics courses

WILLIAM D. MARSLAND, JR.

The Frank 1. Seiler Research Laboratory, U.S. Air Force Academy
Colorado Springs, Colorado

A digital computer is now available at most colleges
and universities in the United States. In those where
there is none today, the advent of time-sharing systems in the next several years should result in a computer capability for all at a nominal cost.
Much effort has been expended and much has been
written on the use of computers in engineering education. To a much lesser extent, efforts are under way
to introduce the use of computers in the behavioral and
social sciences. It seems to me that one very important
and basic area where a digital computer could be very
helpful is being overlooked, namely, to provide increased insight and understanding of many basic areas
of algebra, trigonometry, analytic geometry and calculus, or introductory modern analysis.
Let us consider several examples from the aforementioned areas and see how the computer could be
used as an aid in giving the student a greater insight
and understanding.
In the development of Naperian logarithms one defines:
limo (1

!lEGIN
~I[J

REAL

FO'10,6,"20.10)J
LINE 4(2.1SH
FOil )( • • 00050 STE/' -.OOOOS UNTIL. ,00010. ,00009 srE'
-,0000' UNTIL .000005 DO

"OIlM4T

"ILE
IIEIIIII

• (t • l()eO/lOJ
1111 IT£( LI NE. r. x. [)J

[

END'
END.
0.000500
0.000450
0.000400
0.000350
0.000300
0.000250
0.000200
0.000150
0.000100
0.000090
0.00006'0
0.000070
0.000060
0.000050
0.000040
0.0000)0
0.000020
0,000010

2. 717602S~03
2.71167050e.
2.7177)64016
2.7176062976
2.717G7419S5
2.7179420919
2.71~0099Vl3

2.71110Ta1143
2.11111460SST
2.71615;~571

2.71111730365
2.71111116484&
2.71112005454
2.7111<'140037
2.71112273962
2.7111<'0011571
2. 716~o;55664
2.711126UHS

Figure t.

In studies of probability one is rarely able to generate
empirical data using present-day classroom-teaching
techniques. For example, consider tossing an honest
coin and counting the heads and tails. A teacher might
make an assignment for each student to toss a coin
(assumed to be honest) 100 times and to count the
number of occurrences of heads. Then in class the results could be collected and the overall results discussed.
The teachers would then point out what one might expect if the number of tosses were increased. With the
aid of a digital computer and a pseudo-random number
generator, data for 50 sets of 10, 100, 1,000, and
10,000 tosses can be generated in less than 5 minutes
providing much more data and the opportunity for observation of what happens as the number of tosses is
increased, thus providing the student with the data
needed to gain insight and to draw his own conclusions.
The program given in Fig. 2 assumes the existence of

+ x )1/x == e

A rough calculation of this limit is usually given for
three or four easily calculated points showing that it
does indeed gradually increase above 2.7. The comparatively simple computer program presented in Fig.
1 (written in Burroughs B5500 Extended ALGOL)
can bring out the approach to the limit much more
vividly.
As more display devices become available, the above
results may well be shown even more effectively in the
form of graphical output. At the ACM 1965 National Conference, G. Culler of the University of California described a research project of this type that is
being investigated this fall in mathematics classes.
111

112

Use of digital computers in basic mathematics courses

the pseudo-random number generator. This approach
offers the opportunity for a side discussion of what is
meant by a random number and the problems of
generating one.
In calculus we define the Riemann integral as the
limit of a summation process. The approach of the sum
to a limit for a few specific examples can be demonstrated quite readily using the computer. Consider:
(2 3
) 0 x dx

X4]2
4 0

A computer program using rectangles to approximate
the area under the curve is given in Fig. 3 for the cases
using 2, 4, 10, 100 and 1,000 rectangles.
The latter example also serves as a natural intro-

duction to methods of numerical integration, a topic
frequently overlooked in calculus courses, but essential in many practical problems.
Two of the examples given above have been related
to the convergence of some process. In many calculus
courses the convergence of series is discussed, but the
student is normally left with the impression that if one
can prove that a series converges, the goal has been
achieved and everything will work out all right. Unfortunately, even with the fastest computers presently
available, one frequently cannot afford the luxury of
using enough terms of the series to compute desired results. Rates of convergence should also be discussed in
calculus and can be nicely demonstrated with the aid of
the computer.

IUIII
LIIIE , (hUH
I. J, K. COU"T ••• 'UII •
MEAII ,
".'0.".116). 1".60. -"UN. -. '20,10). MCUO.
-"0, 0' TOUES-. 113. -NO. Of' MUO'-I)'
CO .... ENT
IT 15 ASSUM[O TMAT TME CO"'UTU CENTU MU A UlIoOII NUll''''
I[N[IUTOII AYAlUIlLE. TME CODE 'Oil SUCM A IUENATOII SMOULO IE
IIISEllTEO MEllE. SEE AltiOIlJTMII 'fUM8tll IU ·1" T.. E COMMUNICATIONS 0'
Tift ACII 'Oil AN [UII/'LE 0' A UIIOOM NUMSER fiENUATOII. •
'Oil J • 10. 100. 1000. 10000 Oil
1£lIN
""!TULI"E. II)J
SUM. 0'
'Oil I • I ITEI' I UNTIL 50 00
"EGI"
COUNT. 0.
'Oil K • I Sf'I' t UNTIL J 00
1£1111
• • IIAltoOllCO.loOH
I' • •
TMEN COuNt. COUNT. I •
EIIO.
""ITE CLINE. ,. J. COUIITU
SUM • SUM • COUitT ,
[NO'
II[AN • SUMI'O •
""!TECLINE. G. ItUII)J
"" ITE lL IN[(I'ABU)J
EIIO'
[NO.
NO, Or TOIIU
Nu. 0, TOUU
itO, 0' NUOI
,itO. 0' NUDI
'ILE
IIIT£I'"
ilEAL
'OllMAT

•• 0.

lOSSts

10
10
,0
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10

10
to
10
10
10
10
10
10
10
10
10
10
10
10

10
10
10
10
10

10
10

10
"EAN •

NO. 0' MUDS

,
,,
,
3

5
4

•,
I
S
4
5
5
7
3
3
I

•
,•,
,
•
,•
5
I
5

S
3
3

••
,
,
,
6

7
7
3

,
S. )200000000

100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
tOO
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100

M,AN •

1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000

"
.,
52

,2
51
U

40
50
U

51
52

.,
"

.,.
"
..
"
..."
U
40

53
,7

.""
.
54

52
U

U
51

61
50
52
U
.5
52
58

M,AN •

... 6000000000

Figure 2

522

Sl'
4111
4"

511
503
U3
5U
492
'03
'OS

!H2
511

41'

'Ill
49 •
50'

.,.

..S
49S
507
.,0

...

.93
Stl
507
497

'"
5tJ
410
'50

'S'
41'
51t
50'
500

...'"
'"

'92

500
412
491
."

.n.

. .0
50'
'000000000

11:.1. 0' TOSSEs
10000
10000
10000
10000
10000

NO. or

4984
4951
'99T

luOOO
10000

S007

1~000

HI>I>

10000
10000
10000
10000
111000
10000

4986
'943
SOOO

~~~~~

111000
10000
10000
10000
10000
1(1000
10000
111000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000
10000

M,AN •

~£Aj$

500:
'9TO
SOTI

~098

SOl ~

~: 00
5026

SOuS

so:;:

4Hl
4977
4069
~ns

S022
'992
SOU

4961
S005
SOlS
.991
SO~T
50~ 1

5~61

49:;0

5057
SO)2

soa

SOy5
;C:'l
'97&
50~5

SO!).
4l~S

5018

4957
4927
SO 59
49~)

5027
S004
49411

.,,'

1007.6600000)00

Use of digital computers in basic mathematics courses

8E""

~~:~1~ ~::t~~!

r:U.!i). (HlC5!i. "ND. OF RECTUQlU",
112 •. "AREA 9r RECUNGlU"'I)'
I. J •
lI. tlEIGHT, AREA [OII022U
TOTALUEA, STEPSUE •

INTEGER
AIIRAY
REAL

~~~T~(~I:~':!·tO. too. tODD 00
IIUIN

8EGIN

TOTlLAREA • O'
ST[PSIlE • 1.10
Xt OJ. 01
FOR J • t STEP I UNTIL I

X[JJ • X(J-IJ • STEPSIZE 1
tI[IGHT(JJ • X[J)*3J
AR[A(JJ • HEIGHT(J) • ST[PSIZ[ •
TOTALAREA • TOULAREA • AR[A(J·I •

END. IIRJTE(LINE. r, I. TOTALUEA,"
END'
END.

NO. OF RECTANGLES

10
100
1000

AREA

REeTANllES

9.00000
6.25000

_.a-OOO

_.0110_0
_.00600

Figure 3.

The examples selected have been purposely so simple
that with an ample number of time-sharing consoles
around a campus, an instructor might easily assign
one for students to program and reasonably expect the
students to have the results at the time of the next class
meeting. There are many other examples which are
equally simple. By asking the student to explore the
solutions for himself, I believe he will gain much
greater insight and understanding of the processes which
he now accepts frequently on faith and without much
appreciation of the basic concepts.
With the advent of time-sharing systems I expect to
see at least one very simple language or a very simple
subset of a more sophisticated language on most systems. I believe it will be very commonplace to find that
the basic computer courses will gradually migrate to
the freshman year. I feel the time is at hand for
mathematics departments to consider how the computer
can serve them in providing increased insight in the
topics they are presenting. However, mathematics departments frequently make little or no use of computers
and it may often be necessary for other computeroriented faculty members to guide and assist the
.mathematics faculty in the use of computers.
Some other areas where I feel a computer could be
a useful tool are given below. The list is not intended
to be all inclusive and you are invited to augment it
with your own examples.
Demonstration of the solution of equations representing radioactive decay.
Approximation of the zeros of a function.

113

Demonstration of finding the delta for a given epsilon
in continuity discussions.
Proving trigonometric identities (this would be a very
sophisticated application).
Solution of selected transcendental equations using
iteration techniques.
Numerical computations using De Moivre's theorem.
Tests for parallelograms, straight lines, and various
special kinds of triangles and polygons.
Generating empirical data for various probability
studies.
Empirical data on arithmetic and geometric progressions, nth term of a progression, etc.
Demonstration of why one cannot prove theorems by
example instead of by mathematical induction.
Discussion of the need for a hierarchy of operations.
Generation of binomial coefficients, Pascal's triangle.
Demonstration of various limit theorems; some false
conjectures could be thrown in for testing using
the computer.
Demonstration of convergent and divergent sequences and series.
Determination of coincident and parallel lines.
Solution of simultaneous equations, including cases
where the coefficients are measured quantities
with the possibility of limitations on the measuring equipment.
Synthetic division techniques for roots of polynomials.
Checking for rational roots of an equation.
Slope of the tangent to a curve at a point as a limit
process.
Evaluation of a function near a maximum or minimum point.
Demonstration of amplitude period, and phase shifts
for various forms of trigonometric functions.
Demonstration of limit processes arising out of applications of L'Hospital's rule, e.g.,
lim
sin x.
x~O
-x-Verifying definitions of ellipse and parabola as locus
problems.
Demonstration of the derivative of a function as a
limit process .
Using graphical output devices, generate envelopes of
curves by use of derivatives to show usefulness in
curve sketching.
Demonstration of the conclusions of Rolle's theorem
and the mean value theorem.
J ntegration by the trapezoidal process and various
numerical methods.
Development of an algorithm for evaluating an

114

Use of digital computers in basic mathematics courses

n X n determinant.
Evaluation and sketches of hyperbolic functions.
Generating data for polar curves and other parametric representations of curves.
Three-dimensional representations of vectors (sophisticated) .
Least square and other curve fitting methods for

realistic amounts of data.
Examination of maxima and minima of functions
of several variables.
Multiple integrals as limits of double summations.
Demonstration of Fourier series approximations of
functions for varying numbers of terms in the
series.

ERRATA: Final Version of Papers Appearing in
Preliminary Draft Form in Volume 27, Part I

MATHLAB: A program for on-line machine
assistance in symbolic computations *

CARL ENGELMAN
The MITRE Corporation
Bedford, Massachusetts

THE PURPOSE OF MATHLAB
A mathematical scientist experiments. Today, his
test tube and his breadboard are blackboard and
paper. He may, it is true, have available a computer,
but its role is numerical and its results are delivered
not today or tomorrow but the day after the final
programming bug is corrected. The computer is not
present during the most creative phases of the scientist's labor. The purpose of MATHLAB is to provide
the scientist with computational aid of a much more
intimate and liberating nature.
What sort of aid? The basic goal is to provide facilities for those operations which are mechanical. Among
the most common of these are the addition of expressions and equations, the substitution of subexpressions
into a larger expression, differentiation, integration,
Laplace Transforms, multiplication of matrices, and the
solution of. simple equations. Although the greater part
of a scientist's time is spent on these mechanical pursuits
(in fact, an appreciable portion is probably spent in
simply' checking answers and in the eternal bookkeeping problems of getting minus signs and 2 1T'S right),
we must keep in mind that most of the tedious computations associated with the creative aspects of his
work are of a symbolic, rather than a numerical, nature.
If we are to free the scientist from his routine mathematical chores and conserve his energies for the more
properly human activities of interpretation, analysis,
*This project was supported by The MITRE Corporation
Independent Research Program. It was also supported in part,
through access to its computer facilities, by Project MAC, a
M.I.T. Research Program sponsored by the Advanced Research Projects Agency, Department of Defense, under Office
of Naval Research Contract No. NONR-4102(Ol).

117

planning and conjecture, then we must mechanize the
passage from r2 / r to r in addition to that from 1 + 1
to 2.
REQUISITES FOR A MATHEMATICAL
LABORATORY
I should like to outline here the properties I feel
are required of a mathematical laboratory, not in terms
of the range of mathematical operations available, but
rather in terms of its spirit and feel.
1. It should be capable of ordinary numerical computation. This implies the ability to perform
arithmetic, to compute functions or to look up
their values in tables, and to draw graphs.
2. It should be capable of a wide spectrum of
symbolic computations.
3. It should respond to simple user commands.
MATHLAB is intended for the physicist, not the
programmer. The commands should be no more
complicated than his thoughts. If he wishes to
enter an equation into the computer, he should
need only to type the equation in a notation like
that of ordinary mathematics. If he should then
wish to differentiate that equation with respect
to x, he should have to give a command no
more complicated than "differentiate (x)."
4. It must be expandable by the expert. The language, fUI\ctions, and subroutines of the laboratory must be such that it will grow as an organism. If today we write programs for symbolic
differentiation, we should expect, tomorrow, to
employ them in programs for power series
expansions. The opportunity to expand the pro-

118

Mathlab

grams should be open to anyone who masters a
well-defined and common computer language.
5. It should be extensible by the user. Although the
ability of the physicist to augment the existing
programs will no doubt be severely limited compared to that of the programming expert, he
should be provided tools for doing certain simple things for himself, such as changing notational conventions or teaching the machine the
derivatives of his favorite functions.
6. The computer, as viewed by the user, must be
intimate and immediate. The user should have
next to his desk a console consisting of a typewriter or, preferably, a typewriter and a scope.
Economy might, in some cases, dictate the substitution of a plotter for the scope. These are connected to a large, fast, on-line, time-shared
digital computer. He communicates with that
computer by typing messages on his typewriter or
by means of a light-pen on the scope. The computer replies by means of the same machines. It
types both messages and equations. On the scope
it displays both equations and graphs. Above all,
the response time to the user's requests must be
short.
BRIEF SUMMARY OF EXISTING PROGRAMS
FOR SYMBOLIC MATHEMATICS
The first program we should mention was written
for Whirlwind I by J. H. Laning and N. Zierlerl and
was not really a program for symbolic computation.
But, even though the program was capable only of
numerical computation, it could accept programs written as simple symbolic mathematical expressions and
perform them for a user with no machine language experience.
A later program that could accept instructions in the
form of symbolic expressions, but which was also
limited to numerical computations, was, of course,
FORTRAN. 2
The ALPAK system3 • 4. 5 at the Bell Laboratories, designed by W. S. Brown, J. P. Hyde, and B. A. Tague,
is a program for the manipulation (multiplication,
division, differentiation, etc.) of polynomials and rational functions. It is not a system of symbolic computation in the sense we are using that expression.
Although its instructions are FORTRAN-like, its input/
output, as well as its internal representations of polynomials and rational functions are lists of numbers
representing the relevant coefficients and exponents.
Owing to its entirely numerical nature, the program is
quite rapid.

Another interesting development in the field of numerical computations is the STL On-Line Computer6
designed by G. Culler and B. D. Fried at the Space
Technology Laboratories. Although the programming
facilities are similar to FORTRAN, the instructions are
entered, in an on-line environment, through the pushing
of buttons labeled "+ ," "sqrt," etc. After subroutines
are debugged, it is possible, by pushing buttons, to
define the effect of a short sequence of button pushes
to be the same as any sequence of first-level pushes.
These subroutines of higher-level button pushes can
themselves be incorporated in the definition of button
pushes at a still higher level.
Probably the most fundamental development, to
date, for the adaptation of computers to symbolic computation is the design by J. McCarthy, for just such
purposes, of a language called LISp 7,8 (for LISt Processor). A great majority of the symbolic computation
programs extant are written in this language.
The foremost problem of content facing the construction of a mathematical laboratory today is probably the writing of a satisfactory program for the simplification of mathematical expressions. Simplification
is the unconscious of mathematics. We all simplify expressions every day, making choices appropriate to the
occasion but of which we are almost totaly unaware.
For this reason, it is much simpler to program a more
advanced formal computation, such as differentiation,
which has exact rules of which we are conscious, than
it is to have the machine simplify the answer after it
completes the differentiation. We cannot afford to enter
into a detailed discussion of the current attempts to
solve this problem, but we would like to mention, in
chronological order, the authors of three LISP programs for the simplification of mathematical expressions: T. Hart,9 D. Wooldridge, Jr.,I° and W. A. Martin. 11
A LISP program for symbolic differentiation was
written by K. Maling.l2 Although it suffered from some
weakness of simplification and from its input; output
being restricted to well-formed LISP expressions (which
are not nearly as legible as those written in ordinary
mathematical notation), it was certainly a dramatic
early demonstration of the ability of LISP to handle
formal mathematical computation.
In my opinion, the most impressive example of symbolic mathematics yet performed by machine is the
formal (or indefinite) integration program of J. Slagle,
written as a doctoral thesis under Prof. M. Minsky at
M.I. T.l3 Perhaps the most interesting aspect of this
program is that it was heuristic rather than algorithmic.
It possessed a small table of integrals and tried to re-

Mathlab
duce the problem toone or several found in that table
by the same bag of tricks possessed by a good college
freshman.
The remaining programs as we shall discuss do not,
like those above, represent attempts to perform particular symbolic processes (simplification, differentiation, integration) on a computer. They are, rather,
attacks on the whole problem of building a system of
symbolic computations.
The first such program we should like to mention
is the mathematical laboratory project of W. A. Martin, a work-in-progress as a doctoral thesis at M.I.T.
under Prof. M. Minsky. This work is by far the closest
known relative to our MATHLAB. At present it appears that the main difference of emphasis will be that
Martin's work will stress very broad input/output
capabilities. He is, for example, working on input from
scopes achieved by signifying with a light-pen an interesting sUbexpression of a previously "printed" expression,14 as well as anticipating using the scopes for
handwritten input. The emphasis in our program is
more in the direction of continually increasing mathematical powers. Since both programs are written in
LISP, a good deal of exchange should be possible.
Another approach to the mathematical laboratory
problem is a project, under the direction of L. C.
Clapp,15 while he was employed at Bolt, Beranek, and
Newman. The programs are not written in LISP but
in a much simpler and weaker list-processing language
of Bolt, Beranek, and Newman's own invention, called
SIMPLIST. While this language and program might
serve well for rapid performance of simple computations, e.g., adding two symbolic equations or evaluating
a function entered symbolically, it seems unlikely thai
they will be capable of difficult symbolic procedures,
such as a powerful simplification program or symbolic
integrations. These limitations stem primarily from the
fact that their work is married to the PDP-I, a small
computer.
The next system we should like to discuss is the
IBM FORMAC system. This is an extension of FORTRAN which provides it with the ability to perform
a certain amount of symbolic computation. It is under
development in Cambridge under a group headed by
J. E. Sammet. 16 It is quite capable, possessing such
abilities as simplification, substitution, expansion of
products of sums, factoring out powers of a given
variable, and differentiation. For a number of reasons,
however, it does not seem well suited to a mathematical
laboratory. It is program oriented. One does not give
a simple command, such as "differentiate" but rather
one writes a program. True, the language, like FOR-

119

TRAN, is easy for a nonprogrammer to learn, but it
will not have the universal accessibility of, say, the
simple commands of our MATHLAB. The FORMAC
programs can, of course, be run on-line, but they cannot be run line by line, nor were they intended to be.
It is necessary to write an entire program, for the program has to be compiled as an entity, as does a FORTRAN program (in fact, it employs the FORTRAN
IV compiler). Thus, although a useful tool for symbolic computation when you know at the outset what
you want to do, FORMAC would be quite inconvenient for experimentation.
The final program we would like to discuss before
turning to our own MATHLAB is the FORMULA
ALGOL of Perlis and Iturriaga. In one sense, this
language is similar to FORMAC; namely, the ability
to manipulate formulas, i.e., mathematical expressions,
is embedded in what was previously a purely numerical compiler, in this case ALGOL rather than FORTRAN. Viewed this way, its built-in, as opposed to
programmable, facilities are quite meager: restricted
mainly to formula substitution and some rudimentary
simplification. However, this view17 of FORMULA
ALGOL has been superseded by the creation of a
powerful programming language18 which combines the
numerical facilities of ALGOL not only with some
ability to manipulate mathematical expressions but also
with the ability to create and manipulate list structures. In this role of combined algebraic and list-structure compiler, FORMULA ALGOL stands as a predecessor of LISP 2,19 the type of language in which
programs such as MATHLAB will probably be written
in the future.
MATH LAB-ON THE SURFACE
MATHLAB is our current attempt at realizing a
mathematical laboratory of the sort we have been discussing. The program, which has been developed on
the time-shared system of Project MAC at M.I. T. and
on the IBM 7030 at The MITRE Corporation, is continually growing, and the following description is accurate as of June 30, 1965. We do not feel that MATHLAB has, as yet, sufficient mathematical powers to
be of aid to a general user, except with respect to special
and occasional problems. It does, however, possess
most of the qualities postulated in the previous section as requisites for a mathematical laboratory..

1. Numerical computations. It is very weak in this
department because we decided at first to study
symbolic computation as it represented the crux
of our problem. It cannot draw graphs or evalu-

120

Mathlab
ate common transcendental functions. We can
evaluate algebraic expressions with numerical
arguments in a variety of ways.
Symbolic computations. Here we can perform
many common tasks. We can simplify, substitute,
add equations, differentiate, integrate a little,
solve equations, etc.
The user commands are simple. If the user has
stored an equation called "e1" and wishes to
differentiate both sides of it with respect to "x"
and call the resulting equation "e2", he need only
type: differentiate (e1 x e2) .
The program can be expanded by any LISP programmer. In fact, we are doing this all the time.
MATHLAB can be extended a little by the user.
He can teach the machine the derivatives of functions and change the names of system commands.
It is intimate. The user types in some initial
equations; the computer acknowledges them.
The user requests certain symbolic manipulations;
the computer performs them and types back the
answer. The user types in some expressions or
numbers and requests the computer to substitute
thes~ for certain variables in a previous equation;
the/_computer types back the answer, etc.

Mathematical Notation
In this section and the next we wish to describe
MATHLAB as it appears to the user. There are some
minor differences between MATHLAB as it exists at
Project MAC and on the STRETCH at MITRE. Where
such difference3 occur, we shall describe the situation
at MITRE. First, what sort of expressions may a user
type to denote mathematical quantities? The answer is:
those expressions, composed in the ordinary way, of
the following entities:
1. Numbers: 1, S/2, 2.S
2. Words, representing symbolic variables, composed of strings of letters and digits, the first of
which is not a digit:
x, distance, xl, xlsub2.
3. Operation symbols:
+ (addition)
- (subtraction or minus)
* (multiplication)
1 (division)
t ( exponentiation)
4. Parentheses: ( and ). These have two functions.
The first is to ensure the desired interpretation
of certain expressions, e.g., to distinguish S*

(x + y) from S*x + y. The second use of parentheses is for functional notation, e.g., sin(x).
S. Comma: The comma, besides its end-of-message
role to be discussed shortly, serves to separate
the arguments of a function of several variables,
e.g., f(x, y, z).
All blanks are ignored. The rules of precedence of
the operational symbols are conventional (FORTRAN)
except that, in the absence of parentheses, etxt2 denotes et(xt2), not (etx)t2 == et(2*x).
For example, if the user wishes to enter the mathematical expression (in conventional notation): sin(Sx
+ y2), he need only type (in MATHLAB's notation):
sin(S*x + y 2).

The System Commands
The program as we have developed it accepts three
types of symbolic quantities, called expressions, functions, and equations, which are stored in the computer
and which can be referred to and manipulated by
name. An expression is a mathematical quantity or
expression referred to by its ( one word) name. A
function is a mathematical expression and an ordered
list of bound or dummy arguments and is also referred
to by a (one word) name. An equation is really two
mathematical quantities, one for the left and the other
for the right side of the equation, referred to by a (one
word) name. The initial expressions, functions, and
equations in an experiment are entered into the computer by a function called "denote". For example, one
might type:
denote nil
d == 1/2*a*tt2,
el ==== rt2 == xt2
f (x,y) == x + y,

+ yt 2,

The first three commas signify the end of individual
definitions and the fourth comma tells the computer
that this is all the information we choose to give it
for the time being. The word "nil" is vestige of the
hidden fact that "denote nil" is really a couple for the
LISP evalquote operator. The effect of this input is to
store in the computer an expression whose name is d
and whose value is 1/2*a*tt2, an equation whose name
is e1, whose life side is rt2, and whose right side is
xt2 + yt2 and a function whose name is f, whose list
of dummy arguments is (x y), and whose value is
x + y. From this point on the user never again has to
type in rt2 == xt2 + yt2 but can simply refer to el.

Mathlab

Incidentally, the response to the above instruction (we
shall, from now on, use the convention that we speak
in lower case and the computer in capitals) is:
THANKS FOR THE EXPRESSION D, THE
EQUATION El, AND THE FUNCTION F.
In terms of these basic constructs of expression,
function, and equation, we shall describe the various
system commands.
repeat (x)
This command repeats x to the user. x may be the
name of an expression, a function, or an equation. The
format for an expression or a function is similar to
that of denote. For an equation, if x were el above,
then "repeat" would print:

(El) Rj2

== Xj2 + Yj2

This same format is used when any of the succeeding
commands is called upon to print an equation.
Pleasesimplify(x y)
This command simplifies x and names it y. In this,
as in the following commands, the name "y" maybe
the same as the name "x". In that case, the old x is
lost. If "x" is the name of an equation, both sides are
simplified independently. For a detailed discussion of
the scope of simplification, the reader is referred to Ref.
10, which discusses the simplification program we employ.
forget (x)
If x is a complicated expression, function or equation, its storage might be burdensome. The command
"forget" allows the user to retrieve the space when it
is no longer needed.
substitute ( (vI v2 ... vn) x y).
The first argument "(vI: ..vn)" must be a list of
names of expressions and functions in any order. The
value of each "vi" is substituted in x at each occurrence
of the name "vi". The new equation, expression, or
function (depending on x) is named "y".

At this point we should like to state more precisely
the meaning of the denote and substitute instructions.
Should we give the command:
denote nil
z == x + y,
x and z would be quantities of quite different natures.
We shall refer to "x" as a formal symbol,' it is without
meaning. "z", on the other hand, is the name of an
official expression; its meaning, which we normally
refer to as its "value", is the quantity x + y constructed
of the formal symbols x and y.

121

If we now type the instructions:

denote nil
x == 5*t,
an expression whose name is x and whose value is 5*t
would be created, but this would in no way affect the
status of x in the value x + y of the expression whose
name is z. That x remains a formal symbol. The fundamental connection that can be established between these
two occurrences (of different types) of the character
"x" is through the instruction "substitute".
If we now type:
substitute ( (x) z w)
the program will look for an expression whose name
is x, find that its value is 5*t, look for an expression
whose name is z, discover that its value is x + y, substitute the quantity 5 *t (containing the formal symbol
t) for all occasions of the formal symbol x in x + y
(obtaining the quantity 5*t + y), simplify this to 5*t
+ y, and create a new expression whose name is wand
whose value is 5*t + y.
Substitutions take place only upon command, never
automatically. This is as it should be. The user may
have previously informed the computer that x ==
r*cos (t), but he might like to type x without having it
automatically changed to r*cos (t). Automatic substitution schemes are not only undesirable, but are also
prone to interminable loops.
The substitution of a function in a function is defined
recursively so that it will operate to any depth. For
example, if
g(u, v) == f(f(u, v) ,f(u, v»
and
f (x, y)

== x + y,

then the command:
substitute( (f) g h)
would yield:
H(U, V)

== 2*U + 2*V .

It is probably better to think of substitute in this cir-

cumstance as a command to unwind the functional
definitions.
There is an important restriction on the substitution
of expressions within functions. One may only substitute
for the unbound variables or parameters of a function.
For example, if f(x) == x + t, one may substitute for
t but not for x. It is expected that the casual user will
attempt to violate this rule and instructive error messages have been prepared. This restriction is dictated
by the meaning of a dummy variable of a function, but

122

Mathlab

any desired result can surely be achieved by a short
sequence of commands. In this case, we could, for
example, denote hey)
f(x) and then substitute first
f and then x in h.
This description may seem too detailed, but an understanding of the distinction between an expression and
a formal symbol as well as the function of the substitute
instruction is fundamental to an understanding of
MATHLAB. We shall presume the extension of these
concepts to equations and to other commands, e.g.,
differentiate, is apparent.

respond: (E2) D
l/2*A*Tt2. Should x be a function, the effect is best described by an example. Employing the function "f" of denote above, "makeequation
(f e3)" would yield: (E3) F(X, Y)
X + Y.

add«ql q2 ... qn)name)

e must be an equation whose left side looks like
"f(xl, x2, ... ,xn)" with all the xi simple (one word)
formal symbols. We get a new function "f" whose
dummy variable list is "(xl x2 . . . xn)" and whose
dummy variable list is " (x 1 x2 ... xn)" and whose
value is the right side of e.

The q's can be equations, functions, expressions, or
numbers in any order. If there is at least one equation
among them, name is an equation; if not and some q
is a function, name is a function; otherwise it is an expression. Equations are added by adding left sides and
adding right sides independently. Expressions, functions, and numbers are added to an equation by adding
their values to both sides of the equation. If functions
are added to form a new function, the list of dummy
variables of the new function is the union of the lists of
dummy variables of the old functions.
mUltiply ( (ql. .. qn)name) . ..

Similar to above.
subtract(x y name) . ..

Similar to above, but only two equations, expressions,
functions, or numbers are subtracted instead of an indefinite number as in add and mUltiply.
division (x y name) . ..

Similar to above.
raise(x y name) . ..

Similar to above. name = xY•
negative(x y) . ..

y=-x.
invert (x y) . ..

= l/x.

//ip(x y) . ..

x must be the name of an equation. y becomes the
name of that equation with the left and right sides
interchanged. This is useful if we wish, say, to add the
left side of one equation to the right side of another.
makeequation (x y) . ..

x must be the name of an expression or function. If
x is an expression, an equation is formed whose name
is "y", whose left side is the name "x", and whose right
side is the value of x. For example, using the expression "d" of denote above, if the instruction: "makeequation (d e2)" were given, the computer would

makeexpression ( e ) ...

e must be an equation whose left side is a single word.
Then an expression is formed whose name is the left
side of e and whose value is the right side of e. For
example, "makeexpression(e2)" would now produce:
D
112 *A *Tt2, which is where we started.

make/unction (e) . ..

The purpose of these three "make" commands is
twofold. The first purpose is to guarantee the legality of
certain succeeding commands. In the example occurring
in the description of "makeexpression" above, e2 could
not be substituted in another equation but d could. The
second is to ensure the accessibility of certain results.
E.g., taking e2 and f as above, compare the effect of
the two following sequences of commands:
First:

add ( (e2 f) e)
(E) D + X + Y== l/2*A*Tt2

+X +Y

Second:

makeequation (f e101)
(El01) F.(X, Y) = X
Y
add«e2 elOl) e)
(E) D + F(X, Y) == 1/2*A*Tt2

+X +Y

expand ( (xl x2 . .. xn» . ..

This produces no immediate result but affects all
succeeding simplifications. Whenever one of the x's
(which are formal symbols) occurs in a product of
sums, that product is multiplied out. The following
dialogue will clarify this.
denote nil
e3 ==== yt3 + y*y == (x + z) * (u + v),
THANKS FOR THE EQUATION E3
pleasesimplify (e3 e4)
(E4) Yt3 + Yt2 == (X + Z)*(U + V)
expand ( (x u))
YES
pleasesimplify (e3 eS)
(ES) Yt + Yt2 = X*U + X*V + Z*U + Z*V
The "expand" command affects not only the command "pleasesimplify" but other commands, such as

Mathlab

123

"substitute" or "add", which always simplify their answers.

derivative of arctan for future use.

factor( (x y . .. »

v must be an expression which is a rational function
of x with (rational) numerical coefficients. w is its
indefinite integral. For a more precise discussion, see
the following section. We give an example. If:
v = (x + 1)/( (xt2 + 1)*(xt2 + x + 1)t3),
then the command
integrate(v x w)
yields the result:

...

Like "expand", this produces no immediate result
but affects all future simplifications. It causes the collection of all terms containing, as a factor, a power of
x and similarly for y. The order of formal symbols in
the list "( x y... )" implies precedence. In this case, the
term "x*y" will count as y occurrences of x rather than
x occurrences of y.
We give an example:
denote nil
mitre = a*x + 2*x + x*y + y + b*y + 4*x
t2 + c*xt 2,
THANKS FOR THE EXPRESSION MITRE.
pleasesimplify (mitre bedford)
BEDFORD = A*X + 2*X + x*y + Y + B*Y
+ 4*Xt 2 + C*Xt 2
factor «x y»
YES
pleasesimplify (mitre bedford)
BEDFORD = (2 + A + Y)*X + (4 + C)*
Xt 2 + (1 + B)*Y
By calling the functions expand and factor at different
times with different arguments, the user can maintain
control over the form of his answers.
differentiate (y x yprime) . ..

Differentiates y with respect to x and calls the resulting expression or equation (dependjng on y) yprime.
At present, all differentiation is explicit,' i.e., a term is
considered as constant in x, unless x appears in it
explicitly.
learnderivative . ..

Allows the user to teach the computer the derivative
of a new function. The format of this command is best
explained by an example.
learnderivative nil
arctan,

x,
1/(1 + xt2),
The need for this command has, to some extent, been
obviated by an improvement in the differentiation program. If the computer is asked to differentiate arctan
(x)t2, it will decide that the answer is twice arctan(x)
times the derivative of arctan (x) . If it should then discover that it does not know the derivative of arctan, it
will try to obtain it from the typewriter. If it succeeds,
it will complete the differentiation and remember the

integrate (v x w) . ..

W = (2/3*Xt3 + 1/2*Xt2 + 1/3*X -

1/2)/
(Xt4 + 2*Xt3 +3*Xt2 + 2*X + 1)
+ 1/2*LOG(Xt2 + 1 - ARCTAN(X) 1/2*LOG(Xt2 + X + 1)
+ 7/(3*SQRT(3»
*ARCTAN«2*X + 1)/SQRT(3»

solve(e x) . ..

e must be an equation that is equivalent to a rational function (with symbolic coefficients) of x. At
present "solve" can handle only those equations which
are really (although not necessarily explicitly) quadratic or linear in x. The reply to this command, excepting those cases which the program cannot solve, takes
one of three forms depending on whether the computer analyzes the equation to be linear, quadratic with
distinct roots, or quadratic with a double root. The
following three examples illustrate these different responses.
1. If e is the equation: y = a *x
(e x)" yields:

b, then "solve

THIS EQUATION HAS A SINGLE ROOT.
X= (Y-B)/A
2. If e is the equation: a*xt2
"solve (e x)" yields:

+ b*x + c =

0, then

THIS EQUATION HAS TWO ROOTS, NAMED
FIRSTROOT AND SECONDROOT.
FIRSTROOT = 1/2*(- B
+ SQRT(Bt2 - 4*A*C»/A
SECONDROOT = 1/2 * ( - B
- SQRT(Bt2 - 4*A*C»/A
PLEASE RENAME THOSE WHOSE PRESERVATION YOU WISH TO ENSURE.
3. If e is the equation: xt2 - 2*b*x + bt2 = 0,
then "solve (e x)" yields:
THIS EQUATION HAS A DOUBLE ROOT.
X=B
We give a fourth example which, though simple, exhibits more of the meat of the program:

124

Mathlab

4. If e is the equation: 1/ (xt2 - 1) = 1/ (x - 1),
then "solve (e x)" yields:
THIS EQUATION HAS A SINGLE ROOT.
X=O
There are two points here. First, the equation does
not appear, at first glance, to be linear but the program analyzes it as such. Second, a naive attempt at
solving the equation, e.g., inverting both sides, could
yield what we feel is an extraneous root at x = 1. For
an explanation of how the program avoids this trap,
see the following section.

rename (x y) . ..
The expression, function, or equation that had the
old name "x" obtains the new name "y."
newname(A B) . ..
This command differs from all previous commands
in that it affects, not the data, but the system commands themselves. It creates a new system command
B whose effect is identical to A and which exists in
addition to A. For example, if the user tires of typing
"differentiate," he can give the command:
newname (differentiate d)
after which the command:
d (y x yprime)
will have exactly the same effect as
differentiate(y x yprime).
Before digging beneath the surface of MATHLAB,
it might help clarify some of the preceding if we give a
very short sample session possible today.
denote nil
el == rt2 = s*t,
s = xt2 * y,
t = log(w)/x,
THANK YOU FOR THE EXPRESSIONS S T
AND THE EQUATION El.
substitute «s t) el e2)
(E2) Rt2 = X*Y*LOG(W)
denote nil
w = sin(xt 2 + yt2),

denote nil
THANKS FOR NOTHING.
This last crack on the computer's part is indicative
of the fact that most of our commands have very heavy
error protection. If the user makes a mistake in constructing an expression or an equation or tries to give
an expression a name already assigned to an equation,
etc., he will receive an instructive error return.
MATHLAB-BELOW THE SURFACE
LISP

The entire MATHLAB program is written in LISP. 8
The system commands are all addressed to the LISP
evalquote operator, e.g., the command "differentiate
(h t dh)" presents the evalquote operator with a couple
consisting of the function "differentiate" and the list of
arguments "(h t dh)I".

Numbers
The LISP system, written by R. Silver and P. Heckel20
for the IBM 7030 (STRETCH) at MITRE, contains
only one type of number, namely, rational numbers,
i.e., ordered pairs of integers. If any of the numbers,
12/5, 24/10, or 2.4, is typed in, it is converted to the
rational 12/5.
In addition to rationals, MATHLAB possesses another type of number: the rational power of a rational
number. It will, for example, compute the value of
(81/16)t(2/3) to be 9/4* (3/2)t(2/3).

Internal Representation 0/ Mathematical Expressions
The internal representation of any mathematical expression is a well-formed LISP S-Expression in a prefix
notation. If the user types:
denote nil
sin(pi*t),
v = tt2

then there is stored on the property list of the atom
"v" the property EXPRESSION followed by the SExpression:
(PLUS (EXPT T 2) (SIN (TIMES PI T»).
Should the user then type:

THANK YOU FOR THE EXPRESSION W.
substitute « w) e2 e3 )
(E3 (Rt 2 = X*Y*LOG(SIN(Xt2 + Yt2»
differentiate (e3 x e3prime)
(E3PRIME) 0 = Y*LOG(SIN(Xt2 + Yt2»
+ 2*Xt2 *Y*COS(Xt2 + Yt2 )
/SIN(Xt 2 + Yt 2 )

differentiate (v t dv)
there is stored, upon completion of .the differentiation
and simplification, on the property list of the atom
"dv" a pointer to the S-Expression:
(PLUS(TIMES 2 T) (TIMES PI
(COS(TIMES PI T»).

Mathlab
This is translated back into the original infix notation
and the typewriter prints:
DV

= 2*T + PI*COS(PI*T)

Equations are stored similarly to expressions, except
there are two pointers, one to the S-Expression representing the left side of the equation and one to the
right. Functions are stored by having the indicator
FUNCTION point to the listed pair consisting of the
list of dummy variables and the value of the function
represented by the corresponding S-Expression.
Besides the obvious need for well-formed LISP expressions, there are two reasons for our choice of this
internal representation of mathematical expressions.
First, this representation has become fairly standard
and this allows us to exchange programs with other
workers. Second, the prefix notation turns out to be
well suited to our applications. Consider the differentiation we just discussed. Probably the easiest and
fastest thing for LISP to tell us about any list is the first
item on it: in this case, PLUS. But this is p'recisely the
first thing our differentiation program would want to
know so as to invoke the rule that the derivative of a
sum is the sum of the derivatives. Both the input
(infix~prefix) and the output (prefix~infix) translation programs are written in LISP, the former employing the character-reading functions.
Other internal representations of expressions also
occur, e.g., polynomials as lists of coefficients and rational functions as dotted pairs of lists of coefficients.
All such alternative representations have translation programs connecting them in both directions with the
standard prefix representation.
Simplification and Differentiation

The internal programs for simplification and differentiation have been borrowed from the Stanford Simplify Program. IO They have been modified in two ways.
Simplify has been enlarged to handle a family of sim~
plifications typified by the transformation: (EXPT
(MINUS X) 4/3~(EXPT X 4/3" i.e., (_X)4/3~
x4/3 • This simplification was impossible in the original
Stanford system because 4/3 could only be represente~
by an approximating decimal, such as 1.3333333 and
nothing can be done with (-x) 1.3333333.
Differentiation has been modified so as to look to
the typewriter for the derivatives of new functions. If
necessary, the program will demonstrate, by example,
the correct format for teaching it derivatives.
Integration

The program for the integration of rational functions

125

with numerical coefficients was written by M. Manove2I
at MITRE in the summer of 1964. It is based on a
theorem of Hardy22 that states that the integral of such
a function is of the form Rl
J Rz where Rl and Rz
are rational and R2 has only simple poles. The program
always finds Rl and does the best it can with J R 2 , that
best depending on its ability to factor the denominator
of R2. It is sufficient to consider R2 monic with integral
coefficients. To factor R2, "integrate" first calls a
simple program written by the current author which,
after finding all rational (hence, integral) roots and
factoring them out, will, if left with a quartic, factor it
into the product of quadratics with integral coefficients
(should such a factorization exist). In addition to this
factorization program, "integrate" memorizes the factors
of the denominator of the original problem (if that
denominator is presented in factored form) and uses
these as trial divisors of R2.

"Integrate" is powerful enough to have found several
errors in published tables of integrals.
Solve

"Solve" first brings the equation it has to solve over
on one side. It then combines the various terms into a
single rational expression with one numerator and one
denominator, employing greatest common divisor routines (for polynomials with symbolic coefficients) to
eliminate common factors from numerators and denominators. The roots of the original equation are
then the roots of the numerator of the constructed rational function. If that numerator is quadratic, its roots
are found by the quadratic formula. The vanishing of
the discriminant is the test for a double root. We are, of
course, dependent on the simplification programs here
and within the g.c.d. routines to tell us when complicated S-expressions are equivalent to zero.

CURRENT WORK-September 1, 1965
Current work involves new representations of polynomials and rational functions in several variables with
applications to simplifcation, the factorization of polynomials in several variables, and the integration of rational functions with symbolic coefficients.
In addition, we are working on programs for the
display of mathematical expressions on scopes and the
adaptation of MATHLAB to the AN/FSQ-32 computer at the Systems Development Corporation, Santa
Monica.

126

Mathlab

REFERENCES
1. J. H. LANING and N. ZIERLER, "A Program for
Translation of Mathematical Equations for Whirlwind I," Instrumentation Laboratory, M.LT. Engineering Memorandum E-364 (Jan. 1954).
2. INTERNATIONAL BUSINESS MACINES CORP. Reference Manual 709-7090 FORTRAN Prqgramming System, Prentice-Hall, Englewood Cliffs,
N.J. (1958).
3. W. S. BROWN, "The ALPAK System for Nonnumerical Algebra on a Digital Computer-I," Bell
System Tech. J. (Sept. 1963).
4. W. S. BROWN, J. P. HYDE, and B. A. TAGUE, "The
ALPAK System for Nonnumerical Algebra on a
Digital Computer-II, ibid (Mar. 1963).
5. J. P. HYDE, The ALPAK System for Nonnumerical
Algebra on a Digital Computer-III,ibid (July
1964).
6. B. D. FRIED, "STL On-Line Computer, Volume I
-General Description," Space Technology Laboratories, 9824-6001-RU-000, Redondo Beach,
Calif. (Dec. 28, 1964).
7. J. MCCARTY, "Recursive Functions of Symbolic
Expressions and Their Computation by Machine,"
Communications of the A CM (Apr. 1960).
8. "LISP 1.5 Programmer's Manual," M.I.T. Computation Center and Research Laboratory of
Electronics (Aug. 1962).
9. T. HART, "Simplify, Artificial Intelligence Project," M.1. T. Computation Center and Research
Laboratory of Electronics, Memo 27.
10. DEAN WOOLRIDGE, JR., "An Algebraic Simplify
Program in LISP," Stanford Artificial Intelligence
Project, Memo No. 11 (Dec. 27, 1963).
11. WILLIAM A. MARTIN, "Hash-Coding Functions of
a Complex Variable," M.I.T. Project MAC Memorandum MAC-M-165, Artificial Intelligence
Project, Memo 70 ( June 25, 1964).

12. K. MALING, "The LISP Differentiation Demonstration Program," Artificial Intelligence Project
Research Laboratory of Electronics and M.I. T.
Computation Center, Memo 10.
13. JAMES ROBERT SLAGLE, "A Heuristic Program that
Solves Integration Problems in Freshman Calculus,
Symbolic Automatic Integrator (SAINT) ," Ph.D.
thesis, Mathematics Department, Massachusetts
Institute of Technology, 1961.
14. WILLIAM A. MARTIN, "Syntax and Display of
Mathematical Expressions," M.I.T. Project MAC,
Memorandum MAC-M-257, Artificial Intelligence
Project, Memo 85( July 29, 1965).
15. LEWIS C. CLAPP and RICARD Y. KAIN, "A Computer Aid for Symbolic Mathematics", Proceedings
of the Fall Joint Computer Conference, Spartan
Books, Baltimore, 1963.
16. J. E. SAMMET and E. R. BOND, "Introduction to
FORMAC," IEEE Transactions on Electronic
Computers (Aug. 1964).
17. A. J. PERLIS and R. ITURRIAGA, "An Extension to
ALGOL for Manipulating Formulae," Communications of the ACM, (Feb. 1964).
18. Aa nU. Perlis, Renato Iturriaga, and TOMAS A.
STANDIS, "A Preliminary Sketch of Formula
ALGOL," Internal Report, Carnegie Institute of
Technology, Pittsburgh (Apr. 9, 1965) .
19. R. W. MITCELL, "LISP 2 Specifications Proposal,"
Stanford Artificial Intelligence Project, Memo No.
21 (Aug. 19, 1964).
20. P. HECEL, forthcoming MITRE Aechnical Report (untitled).
21. M. MANOVE, "Integrate: A Program for the Machine Computation of the Indefinite Integral of
Rational Functions," The MITRE Corporation,
TM:"04204 (Apr. 22, 1965).
22. G. H. HARDY, The Integration of Functions of a
Single Variable, Cambridge University Press
(1958).

A time- and memory-sharing executive program
for quick-response, on-line applications
JAMES W. FORGIE

Lincoln Laboratory, * Massachusetts Institute of
Technology, Lexington, Massachusetts

BACKGROUND
The TX-2 Computer, an experimental facility at
M.I.T. Lincoln Laboratory, has been in operation
since 1960. 1 Never a service facility, the computer has
been used principally in a number of long-term research projects that have taken advantage of the special
input/ output capabilities and the direct accessibility of the machine. These projects have included
graphics,2,3 waveform processing/,5 and pattern recognition.6,7 Most of the work on the computer has involved real-time inputs, interaction with output displays, or both. The computer has always been used
as an on-line facility with the bulk of its time allotted in
sessions of several hours duration. Programming has
been in machine language, augmented in the past by
a number of personal macro languages and recently
by a more general macro language for list processing
(CORAL). An on-line macro assembler, MK 4, has
been used both as an assembly program and on-line
operating system by most users.
In the fall of 1963 it was decided to realize on TX-2
an experimental operation-oriented, on-line system in
order to study man-machine interaction in problem
solving, This system would allow the scientist or engineer to make use of the computer throughout his work
on a data-analysis problem without concerning himself with many of the details ordinarily involved in
programming a computer. 8 The system would be based
on a library of computational and display routines that
could be called directly by the user in an appropriate
problem-oriented language. For a problem area in
which library routines existed, it was expected that
single library routines, or short combinations of

routines, would suffice for a high percentage of the
operations he would need. In order to handle the few
remaining cases, the system would include special and
general compilers which the user could utilize. to create
the occasional pieces of program that he mIght need
to complete the solution to his problem.
It was felt that a person using a system of this sort
would probably spend much more time looking at his
displays and thinking about what to do next, than he
would spend actually doing computations. Economic
considerations then dictated that multiple consoles
should be provided and the computer facilities shared
among these consoles. An executive program ~?i.ch
would handle such sharing of the computer faCIlItIes
and related problems of storage allocation and communication became a central part of the system design.
In January of 1964 a commitment was made to realize
such an operation-oriented on-line system on TX-2.
Since experimentation with the new system would put
pressure on the already full schedule of TX-2, a further
requirement was placed on the design of the new system, namely, that its executive program should allow
for the use of TX-2 in something approaching its accustomed style at the same time that the system was
running. Thus the advantages of time-sharing could be
made available to the projects already using the machine. This paper discusses the design of the executive
system, called APEX, which has grown out of those
decisions.
SYSTEM REQUIREMENTS
The design requirements for the APEX system may
be briefly stated as follows:

1. Time Sharing. The system should time-share the

*Operated with support from the U.S. Air Force.

127

128

A time-and memory-sharing executive program

essential computing facilities among a small number of
consoles (perhaps half a dozen) most of which would
consist of an input keyboard and either an output
typewriter or a display oscilloscope, or both.
2. Fast Response. The on-going activities in graphics,
waveform processing, and pattern recognition all involved the use of interactive displays. It appeared that
response times in excess of one second would seriously
degrade the performance of already existing programs
in these areas. In addition, the proposed experiments
with the operation-oriented on-line system called for
the ability to degrade response time in order to measure
its effect on the user. Thus, all proposed applications
of the system called for fast response under at least
some circumstances.
3. Retention of Results. The executive should assume responsibility for the retention of all programs
and data files whose destruction was not specifically
ordered by the user or his program.
4. Subroutine Autonomy. The executive should allow any program, assuming it is written as a closed
subroutine and follows certain conventions, to be run
as an independent program making full use of core
storage addresses and index registers. Routines to be
run in this fashion should be precompiled and stored
in absolute binary form. They should be completely
independent of the routines which call them and thus
might call themselves recursively. The executive should
provide isolation and protection for such routines and
facilitate the passing of parameters to them. This requirement for subroutine autonomy was intended to
achieve speed in the running of library routines by
eliminating the time for compilation or relocation and
was intended to simplify the programming of such
routines by minimizing the number of restrictions they
would have to meet.
5. Flexible Input/Output Services. The executive
should handle the details of all input/output operations.
It should provide continuity for displays and keyboardtypewriter conversation. It should provide for the sharing of I/O devices like printers and magnetic tape
that cannot be duplicated at every console. In so far as
possible, it should leave formats and the interpretation
of commands to the user's own programs.
DESIGN DECISIONS
Early in the design phase of the executive program
a number of policy decisions were made which had a
considerable effect on the· character of the final program. In retrospect the most important were the following;

1. In order to meet the requirements for fast response, memory as well as time should be shared among
the consoles. If the active part of the program for each
console can be kept in core, the time required to
switch between users is greatly reduced. With a small
number of consoles, it appeared that the TX-2 memory
was large enough so that it would often hold all of the
active pieces of their programs, provided the pieces
were of reasonable size. It was therefore decided that
the executive should provide services that would encourage programmers to break large structures into
small units. And to facilitate this sort of memorysharing, it was decided that hardware for relocation
and bounding should be added to the computer.
2. The system should provide program sharing so
that memory sharing would operate efficiently. Large
public routines, such as compilers, should be written as
pure procedures so that they could be shared by all
users. The TX-2 order code allows this kind of program to be written without any special difficulty. It
was decided that the executive should incorporate features to facilitate the operation of pure procedures and
that the hardware necessary to protect them should be
installed.
3. The executive should simulate an apparent computer for each· console. The requirements of the operation-oriented system could have been met by a highly
specialized executive program, but such a design would
not have satisfied the needs of the research projects
that were already using TX-2. Their needs would,
perhaps, have best been served by a time-sharing
system that provided the entire facilities of the computer for each user in turn. Existing programs could
then have been operated in the new system without
significant changes. The large amount of time-dependent interaction between the TX-2 I/O system and
the programs which use it would have required a very
complex executive program in order to allow the
existing I/O routines to operate. The simulation of an
apparent computer similar, but not identical, to TX-2
seemed a reasonable compromise between these two
requirements.
4. There should be no direct communication between
the user and the executive. All input from the user
should be passed through the executive to programs
operating in his simulated computer and translated
there. Commands to the executive would then be passed
back from such a program to the executive. It appeared
both unnecessary and undesirable to tie the system
to any language conventions by building the conventions
into the executive.
5. Insofar as possible, software features should be
realized in programs operating in the simulated com-

A time-and memory-sharing executive program

puters. This decision allows the executive to be as
simple as possible and permits expansion of the overall
software structure without modifying the executive.
6. Compatibility between former TX-2 programs
and programs that would operate in the simulated
computer was not to be a requirement. The design of
the simulated computer should be made to correspond
to TX-2 whenever possible and reasonable. But it was
expected that some changes would have to be made in
all programs to accommodate the input! output characteristics of the executive and to take advantage of the
storage allocation that it provides.
7. Changes in the hardware of the TX-2 computer
were to be considered as legitimate variables in the design work. The computer engineering group was prepared to make reasonable modifications to the computer
when such changes appeared to be the desirable and
economical solutions to the software problems.
Throughout the development of the executive program
there was strong interaction between hardware and
software designs and designers, and major changes
were made in the computer to facilitate the APEX
system. These include the addition of a file memory
( a UNIVAC Fastrand Drum), hardware to trap the
attempted execution of privileged instructions, and
four memory-snatch channels to increase the efficiency
of high speed I/O operations. The most significant
change was the addition of a hardware system called
SPAT (an acronym for Symbolic Page Address Transformation). SPAT, which has been in operation since
January 1965, uses a 1,024 word thin-film memory9
and high-speed transistor circuitry to make a three-level
address transformation within a single TX-2 clockpluse time (O.4p.sec). This transformation makes available to the executive the advantages of paging, segmentation, and complete memory protection. It greatly
reduces the overhead involved in sharing memory
and programs.
BASIC CHARACTERISTICS OF
THE APPARENT COMPUTER
As has already been said, the APEX executive program simulates an apparent computer for each con-sole. The apparent computers may be viewed as somewhat restricted replicas of TX-2 augmented by special
features provided through the executive program. The
core storage for each apparent computer is bounded
and segmented and is limited in total extent to approximately two-thirds of the TX-2 core capacity. The order
code is that obtained by eliminating input/output and
multiple-sequencing instructions from the TX-2 order
code,10.11 and then adding some executive calls to

129

handle input! output, file maintenance, and allocation
of storage in the apparent computer. The number of
index registers is reduced to 15, and some restrictions
are placed on the choice of "configurations" (which
are used primarily to control operations on subwords).
Unlike TX-2, the apparent computer is a singlesequence computer in the current version of the system
(i.e., it has in effect only one program counter), but
the hardware allows, for future expansion to three sequences. In general, programs written for TX-2 will
not operate in the apparent computer, and vice versa.
However,programs which do not involve I/O operations may often be transferred with no change.
The storage structure of the apparent computer takes
advantage of the SPAT address-transformation hardware that has been added to TX-2. The SPAT hardware (which is discussed in more detail in the last section of this paper) breaks core storage into pages of
256 registers, which are organized into books (segments) of up to 32 pages (8,192 registers). The 17-bit
address of TX-2 allows 16 such books to be selected by
the four highest order address bits. Since the total number of apparent addresses exceeds the available core,
some of the books must always be incomplete or empty.
In the apparent computers realized by the APEX
executive program, the user's programs and data are
organized into files. A file is a contiguous group of
registers, which must be some integral number of pages
in length. It always has at least one name, which is
known to the APEX file directory. Files may exceed
one book in length, but they must begin at the start of
a book, and no more than one file may occupy a book.
Executive calls in the user's program determine which'
files are to appear in core at anyone time. A file may
be set up in a book specified by the directory, as is
usually the case for program files, or it may be set up
in an arbitrary book according to the requirements of a
program which is to process it.
All files begin as working storage files with ephemeral names. When a program has finished writing information in such a file, it may give an executive call
to assign a permanent name to the file. Once the file
has been given a permanent name, it will remain in the
file memory until it has been discarded by a call from
the user's program. Thus, all data files that the user
has had occasion to name will be retained from one
session to the next. Files which receive only ephemeral
names are discarded automatically according to simple
rules. Executive calls are available that can give a file
read-only protection, prohibit its use as a program file,
and expand or shrink it by an integral number of pages.
A further feature, called auto-expandability, takes
advantage of the potentialities for dynamic storage al-

130

A time-and memory-sharing executive program

location provided by the SPAT transformation. This
feature allows files which tend to grow with time to
make efficient use of core without putting the burden
of boundary testing onto the programs that build them.
Auto-expandability is accomplished by having the executive expand the file by one page whenever the program makes a reference to an address in the first page
beyond the, current boundary of the file.
The complex of files set up in the user's apparent
memory at anyone time is called a Map. A Map may
be thought of pictorially as shown in Fig. 1 with a
typical setup for a matrix routine. However, a Map may
be equally well described as simply a list of names of
files, together with the number of the book in which
each is to appear in the Map. In Fig. 1, the dashed
lines indicate the potential capacity of each book, while
the solid lines indicate the actual core occupied by the
BOOK
3

BOOK
2

BOOK
1

B B

I ..~'" 'I

I
I

L _ _ _ _ _ _ ..J

IL ______ .JI

4
r------,
I

5
,------,
I
1

I
I

I
I
I
I

I
I
IL ______ .JI

I
I
1

I
I

IL ______ .JI

I
I
I
I
IL ______ JI
12

r-------,I Ir-------,I

.-------,

1
1

I
I

I
I
I

I
I

I
1
I
I
1L ______ JI

15
r-------,
I

'I
I

I
I

1
1
I

I
I
IL _______ .JI

IL ______ .JI

I
I

14
r-------,
I
I

I
1

I
I

1
I
1

I
II

:-----']32 PAGES
I
I
I
I

I
I
I

MAXIMUM

IL ______ .JI
13
r------,

I
I
I

I
I
IL _______ JI

16
r-------,

I
I

I
I
I

L _______ J

I
I

I
I
I
I
IL _______ .JI

Figure 1.

I
I
I

6
r------,

IIL ______ I

I
I

I
I
I
I
IL ______ -'I

Map for matrix addition.

four files. More will be said about the contents of this
Map after a short discussion of the way the APEX
system handles library routines.
One of the principal design requirements for APEX
was to provide a means whereby library routines (or
arbitrary subroutines written by the user) could be
called into core and operated without any conflict between the core-addressing requirements of the routine
and the program that called it. This requirement is met
by providing a fresh Map for the called routine. When
a program wishes to call a library routine to be operated
in a new Map it does so by issuing a Go Up call to the
executive, passing along the name of the library routine
as a parameter of the call. This process is called "going

up" because the new Map is thought of as being put
on top of the Map that contained the calling program.
The reverse process, returning to' a lower Map, is
called "peeling back."
Since there are usually some parameters' that must be
passed to a library routine, a special file called the
Connector is provided. It is common to all Maps and
is used primarily for communication between them, but
it may also be used for small amounts of working
storage by the library routines, which are generally
pure procedures. By convention, the first register of the
Connector indicates the beginning of free storage in the
file. Before issuing a Go Up call, the calling program
normally stores a block of information into the free area
of the Connector and moves the free storage pointer
appropriately. This block contains a Peel Back call
followed by the parameters to be passed to the library
routine. The location of the Peel Back call and the
name of the library routine are then given to the executive as parameters of the Go Up call. On receiving the
Go Up call, the executive produces an entirely new
Map that contains only the library routine and the
Connector and then passes control to the library routine
in the new Map. The routine is entered as though it had
been called as a closed subroutine from a location just
before the Peel Back call in the Connector. It then finds
its parameters in the Connector, inspects them, and
gives executive calls to set up such other files as it may
need to carry out its mission.
Figure 1 shows the state of a Map in which a library
routine for matrix addition has just finished its work.
The Map began with only the file of matrix routines
and Connector file set up. The addition routine found
the names "A" and "B" as its input parameters, set
up the files having those names in Books 1 and 2, set
up an ephemeral file in Book 3 to receive the sum,
computed the sum, and gave the output file the permanent name "C," which it found in the Connector as the
name to be given to its output.
With its operations complete, the routine will now
make a standard subroutine exit, transferring control
to the Peel Back call, which will cause the executive to
discard the new Map and return control to the calling
program on the lower Map at the location just beyond
the Go Up call. If the library routine was to have generated some output which was to be returned directly
to the calling program, the calling parameters would
have specified a location in the Connector into which
the output would have been placed.
This way of handling library routines has a number
of advantages. First, the library routine is written as an
ordinary closed subroutine and is not itself concerned
with going up or peeling back, unless it needs to call

A time-and memory-sharing executive program
another routine in the course of its operation. It may
therefore be operated either by going up to a new Map
or by setting it up in the same Map as the routine that
calls it and then using it as an ordinary subroutine. The
latter mode of operation has speed advantages, but is
limited to situations where the programmer has determined that core assignments and index register usage
are compatible. A second advantage comes about because the ability to change Maps is available not only
to library routines, but also to arbitrary programs
written by the user. The stack of Maps aids the programmer in putting together large, complicated structures, which may exceed both the real core and the
core addressing capacities of the machine. However, he
must keep in mind that changing Maps involves a bookkeeping overhead in the executive and will involve swapping memory (at disc speeds) if real core capacity is
exceeded.
MORE SOPHISTICATED USES OF MAPS
Figure 1 showed a Map that resulted from the operation of a simple library routine. Now consider a more
complex Map which illustrates the use of auto-expandable files and shows some advantages of segmentation
in a large program. Figure 2 is a typical Map used by
BOOK

BOOK

3
r------,
I
I
I
I
I

I
I

I.... ______ JI

I
I
IL ______ J I
5

r------,

I
I

I
I
I
I

I
L
_____ J

1
I
I
I
II.. _ _ _ _ _ _ JI

I
I
I
I

IL ______ -'!

I
I
I

I
I

I
I
I
I
IL ______ JI

6
,..-------,
I
1

I
I
I
I

I
I

I
II.. ______ JI

I
I
IL ______ -'I

~ EJI2
CHOPPER

B'3

I
I
IL ______ JI

I
I
I1.. ______ - '1

II.. ______ ...JI

IL ______ JI

I
I

IL _______ JI

Figure 2.

PARSER

I
I
I
I
IL ______ J I

I
I

I.. ______ J I

Map for a compiler.

a compiler realized within a compiler-compiler called
VITAL * (Variable Initialized Translator for Algorithmic Languages), which is being developed to operate
*The VITAL compiler-compiler project was initiated by
L. O. Roberts. It is now the responsibility of J. A. Feldman

and C. D. Forgie.

131

in the APEX system. The figure shows the Map for the
compiler itself, which is made up of 12 files, five of
which are auto-expandable, as indicated by the arrows
extending downward from the rectangles that represent
the files. The Map also contains the Connector, which is
auto-expandable too. It is not in itself an essential part
of some compilers; but it provides a way of entering
and leaving the Map, and is therefore used extensively
in translating and running programs in some languages.
In operation, the Input Pre-Processor (Book 16)
builds a statement in the Working Storage file (Book
7) from characters obtained from the Input Buffer
(Book 15). The Chopper (Book 12), using language
definitions from the List Structure (Book 4), breaks
up the statement into a string of operators, constants,
and symbolic expressions. This string is stored in the
List Structure using the CORAL subroutines in Book
11. The Parser (Book 13) then determines the grammatical structure of the statement string and prepares a
generating string (also stored in the List Structure)
which the Basic Code Generators (Book 10) and
Loader (Book 0) can use to build the Compiled Code
in Book 2.
The Input Pre-Processor, Chopped, Parser, and
CORAL are independent of the particular language
being translated. The definition of the language is determined by the Semantic Tables, Basic Code Generators, and List Structure. Since VITAL allows the
definition of a language to be modified and augmented
during the compilation, the complete VITAL structure
includes another Map called the Meta-Compiler Map.
This Map (not shown) is similar to the Compiler Map
of Fig. 2, except that the Basic Code Generators and
Semantic Tables are those for the meta-language. When
the services of the Meta-Compiler are needed, it is entered by a Go Up call from the Compiler Map.
Figure 2 shows a program structure that exploits fully
the segmentation of storage structure made available by
the SPAT transformation in TX-2. While it would appear from the figure that there are three spare books in
the Map these books are actually committed to providing for the larger files of Compiled Code and List
Structure that would arise in the case of a larger program or a more complex language. If fewer than 16
books had been available, it would have been necessary
to combine files, losing the advantages of auto-expandability· in conserving core capacity and of flexibility in
switching languages. If more than 16 books were available, greater flexibility could doubtless be achieved, but
the bookkeeping costs to the executive would increase
and could become burdensome in simple situations
where only a few segments are needed.
It is often the case that when a program structure

132

A time-and memory-sharing executive program

pushes the limit on the number of books, it is also pushing the limit on available core. In that event the programmer has the choice of asking for a new Map or
changing the contents of the current Map. VITAL uses
the latter method when it must compile a program so
large that the Compiled Code and the List Structure
cannot be contained in the space shown in Fig. 2. Since
the Compiled Code does not need to be in core while
the List Structure is being built or modified, and since
the Chopper, Parser, and CORAL do not need to be
on hand during a simple compilation from a previously
constructed List Structure, VITAL drops the unnecessary files from core when space is limited.
The APEX system also uses Maps to handle interrupts. The user's program may define special Maps
called Ghost Maps, each of which is associated with a
particular source of interrupts. These Maps are called
Ghosts because most of the time they have no real
existence. But when an interrupt occurs, the Ghost
Map associated with the device that caused the interrupt appears on top of the stack of Maps and control
is passed to it.
For interrupts caused by an illegal instruction, a
boundary violation, or an I/O difficulty, a special
HELP Ghost Map is provided; it automatically takes
the user to a fixed, public routine that stabilizes his
current I/O problems, if any, and then sets up a Basic
Translator that allows him to call debugging or other
routines to his aid. Note that in this situation, the
HELP Ghost Map has suspended the operation of his
program and has given him the full use of his apparent
computer to work on his trouble. In addition, the stack
of Maps has preserved all that was known about his
program structure at the time the interrupt occurred.
He may be able to fix the trouble and continue, discarding only the HELP structure at the top of his stack,
or he may elect to start again at the bottom, forgetting
everything about his old structure.
Another example of the use of Ghost Maps is in
processing break-point interrupts. In each of its core
memory words TX-2 has an extra bit called the metabit
that is unaffected by ordinary instructions. A pseudoI/O device in TX-2 will detect the occurrence of a set
metabit on an instruction, a deferred address, or an
operand and will produce an interrupt if the pseudodevice is turned on. This facility can be used for a
variety of debugging purposes, and the APEX system
provides executive services and library routines that
make the facility available in the apparent computers.
Figure 3 shows the Map stack that will appear in one
such application.
.
The figure assumes that the user has logged in and

/
/
/

r-----------

/
METABIT

GHOST

MAP

mill_. 1_1___1_1_

Figure 3.

Map stack for break-point processing.

used the Basic Translator in Map 2 to call a program
called DEBUG*, which will put break-point interrupts
into his program. Map 3, the DEBUG Map, contains its
own translator, which accepts commands from the user,
allowing him to specify where he wishes to put breakpoints and what response he desires when they are
detected. The DEBUG program calls into its own Map
the program files in which the user wants to put b1:"eakpoints. DEBUG then tells the executive that it wants to
use metabits in these files and says where they are to
be placed. (The executive is involved in this operation because metabits do not exist on the file-memory
drum and must be restored if the file is moved out of
core and back in.) The program structure to be debugged may involve many files, but only those that
are to have break-points need be set up at this time
and given special metabit status.
DEBUG then gives an executive call to define a
Metabit Ghost Map, to which control will be transferred
when the pseudo-I/O device detects a set metabit.
DEBUG defines this Map to contain the same files as
the DEBUG Map itself, but specifies a different entrance point. DEBUG then issues a Go Up call that
takes control to the program to be debugged (Map 4),
which runs in its normal fashion until a set metabit is
encountered. At that point the executive activates the
Ghost Map and transfers control to it. DEBUG, running
now in the Ghost Map, makes note of the break-point
and responds to the user in the way he specified. It may
take some action that he has previously specified, or it
may halt operations and await his further commands.
*The DEBUG; program is developing under the
Prof. T. G. Stockham, Jr., of M.I.T.

guidan~

A time-and memory-sharing executive program
In either case, a Peel Back call given in the Ghost Map
can return control to the program being debugged, and
let it continue to the next break-point. In Fig. 3 the
program being debugged is shown as occupying a single
Map (Map 4), but the system works equally well when
a multiple-Map structure is being debugged.
INPUT AND OUTPUT FOR THE
APPARENT COMPUTERS
The basic console that is available to the user of
APEX has a keyboard, an output typewriter, a display
scope of one sort or another, a light-pen, and a few
push-button switches. Since APEX is an experimental
system, there are differences in the makeup of the
equipment available at the various consoles. Some consoles have no output typewriters, others have no display
scope; a RAND tablet12 will be available on one console,
and some will have a connector with 36 output and
36 input wires to which a user can attach special equipment of his own.
The keyboard and the output typewriter operate as a
fully duplexed system with the executive handling the
typing directly. The executive places input characters
into a buffer file accessible to the user's program. As
each character passes through the executive it is checked
against a table of terminators supplied by the user's
program. If the user's program is inactive, or if a GhostMap interruption mode has been specified, the executive
will take appropriate action when a terminator is encountered. Any group of characters may be defined as
terminators. The typing back of keyboard inputs may be
suppressed .when the keyboard is used as an input
device for the display scope.
Hardware and software techniques for the production
of displays on cathode ray tubes are areas in which
much work has been done on TX-2 in the past. This
work has been characterized by an unusually tight coupling between the display and the computer. In the
past,2.3.13 the picture typically has not been generated
from a simple table of the points and lines to be displayed; it has been produced "on the fly" as the program works its way through a complicated ring structure that contains information not only about the points
and lines to be displayed, but also about relations between thf~ objects that the picture represents. 14 One of
the most difficult requirements placed on the design of
the APEX system was that it should allow display work
of this type to continue in an environment of timesharing. 15 The present design appears to be a reasonable
first compromise between time-sharing and tight coupling of display and computer, but it may well have to
be changed in the not-to-distant future as new tech-

133

niques and new requirements develop.
The display equipment that is currently available
consists of scopes at three consoles. All three are driven
in parallel from a shared vector-and-curve generator16
that gets its inputs directly from the computer memory.
Analog integrators are used to generate lines, circles,
and parabolas from the digital information obtained
from the computer. The analog deflection signals are
sent to all scopes, but only the scope for which the
display information is intended receives an intensification signal. One console has a Charactron tube, but the
others have simple cathode ray tubes which require
characters to be generated by the vector-and-curve
generator.
The information that the user's program wants the
executive to display is contained in a ring-structured
file that the user's program may set up in its own Map
for inspection. However, the user's program cannot
write in the file. When changes are wanted, it must request the executive to make them. This restriction was
imposed because the executive's display routine may
run wild if asked to follow an ill-formed ring. The
structure can contain only one type of information about
relations between parts of the picture: it can show what
parts are subordinate to others. This capability is very
important to the kind of display work that has been
carried on in the past. In such work it is important that
the user's program be able to request that the executive
modify a part of the picture, including all subordinate
parts, without rebuilding the whole file. If the user
wants his program to keep track of other types of relations, the program must keep the extra information in
private files of its own.
The rings in the display file specify the order in
which the elements of the picture are to be displayed,
and the display routine in the executive works its way
through the complex, transmitting the relevant data to
the display-generation hardware. A single pass around
the ring defines a "frame," and the display-generator
is time-shared among consoles on a frame-by-frame
basis. The display routine maintains the display on the
user's console even though his program is inactive because of time-sharing or because it is waiting for an
input. A "push-to-see" button is used by some types of
display programs to keep down the load that display
maintenance would otherwise place on the system.
The task of processing signals from the light..,pen is
closely related to the task of generating displays since
the pen is the principle way in which the user communicates to his program information about the ring
structure from which the display is generated. The
light-pen may be used in two modes-pointing and
tracking. In both modes the executive maintains a

134

A time-and memory-sharing executive program

complete record of all light-pen returns in a buffer in
the file of information being displayed. A light-pen
return while in the pointing mode causes the executive
to place in the buffer information from which the user's
program can calculate both the pen position and the
place in the ring structure from which the element seen
by the pen was generated. In many light-pen applications
it is necessary to associate the pushing of a button or
the striking of a key with the pointing operation. This
association is handled by the executive, and the associated character code is placed in the light-pen buffer.
In the tracking mode, the executive displays a tracking cross every 30 msec. If the pen sees any part of the
cross, the executive moves the cross to center it in the
field of the pen and the location of the center of the
cross is placed in the buffer. Smoothing and extrapolation are done in the tracking routine to achieve good
"writing" characteristics for the pen. The processing of
light-pen signals is a high priority task for the executive
since fast responses are essential, especially in the
tracking mode.
The second category of input/output equipment that
is available to the apparent computers consists of devices that must be shared by them. It includes magnetic
tape and paper tape equipment, units for analog inputs
and outputs, and a high-speed Xerographic printer. This
class of equipment has an unusual status in the APEX
system because most of the consoles will be in the
computer room, and the users at those consoles will
have easy access to the common, shared devices. While
these shared devices posed a number of problems to the
designers of the executive, the solutions are too specialized to the nature of TX-2 to warrant further discussion
here.
CONTINUITY OF OPERATION
IN THE APEX SYSTEM
The user for whom the on-line, operation-oriented
system is designed is a man at work on the solution to
a problem. His problem will rarely be solved in a
single session on the computer, and so one of the major
tasks of the executive system is to remember his programs, files of data, and other quantities that he may
find useful in maintaining the continuity of his work
from one session to another. A portion of the executive
maintains a private directory for each user as well as a
public directory which is shared by all users. A number of calls are available to the apparent console computer to allow a user's program to insert items into his
private directory and to inquire about these items and
about items in the public directory.
Items remembered through a directory are identified

by names, which are strings of up to 50 characters. The
characters are restricted to Roman capital letters, Aarbic
numerals, and period. The directory itself is a ring
structure arranged in the form of a tree to give a
logarithmic search-time for names. When a name is
entered into the directory, a unique block is created for
it in the list structure, and the pointer to that name
block is used as a compact and more efficient substitute for the original string of characters. Remembering an item in the directory involves two calls. The
first asks the executive to accept a string of characters
and return the resulting name pointer. The second call
uses the pointer plus the necessary defining information
to establish an association between the name and the
item to be remembered.
The directory can keep track of the following kinds
of items, either directly or by way of the file memory:
1. Files. A file is any contiguous group of memory
registers. As was explained in connection with the storage structure of the apparent computer, the directory
contains information showing whether the file is to be
protected against writing operations, whether it is to be
protected against being executed as a program, in
which book it is normally to be set up (if it has a preferred book), and whether it contains program or data.
If it contains data, the type of data is shown only by
the internal format of the file which is left to the author
of the program who created it. The executive system
must serve such a wide variety of programmers that any
attempt to reach an agreement on a standard classification of data types seemed hopeless.
2. Scalars. A scalar is a single-register quantity remembered in the directory itself. Scalars are useful for
allowing the user to remember single numbers that are
not part of some larger array of data. They are also
useful in allowing public routines to use the user's
directory to remember certain parameters from one
usage to the next. Thus, for example, a compiler (which
is normally a pure procedure) can remember the name
of the last program it was compiling for each user.
3. Entrances. An entrance is a number associated
with a file. A program file may contain a number of
related routines which perform different functions. Entrances can then be used to call these different routines
by entering the program file at different locations. If a
Go Up call is given to the executive and the parameter
on that call specifies an entrance, the file will be set up
(if it is a program file), and control will be transferred
to the location specified by the entrance. Entrances may
also be used with files of data. For example, an entrance
may identify the start of a particular ring in a list
structure.

A time-and memory-sharing executive program

4. References to public names. A reference to a
public name is a device for allowing a user to substitute a name of his own choosing for a public name
that is unsatisfactory to him.
S. File groups. The file group, as the name implies,
is merely a collection of related files. Its existence in
the directory allows the group of files to be brought in
from the drum and set up in memory by means of a
single call. For example, consider the situation when a
general translator is used to translate a particular
problem-oriented language. In addition to the file containing the translator itself, a file of definitions for the
particular language and a file of working storage must
be set up before any translating can begin. Treating the
files as a group allows the executive to get them all into
core and set up before any attempt is made to run the
program.
The directory not only maintains relations between
names and things; it also maintains relations between
names. A Synonym call to the executive allows the
user's program to indicate that a particular item in the
user's directory is to have a second, synonymous name.
A name may have any number of synonyms. They are
added one at a time by the Synonym call and may be
removed one at a time by the Undefine call. If all of
the names have been removed by the Undefine call, the
item itself will be forgotten by the directory and destroyed. There is also a Drop call that allows all the
names to be undefined and the entity destroyed with a
single call. Synonyms are useful for abbreviatIon and
for substitution of parameters. They are handled by the
executive rather than left to particular translators because it is felt that the user regards them as languageindependent relationships that should endure when he
switches from one translator to another.
As a convenience to the user, and for his protection
against system failures, his private directory and all his
files that existed at the start of a session are retained
until the end. If he ends the session in the normal way
by logging out, his starting directory and any obsolete
files are discarded. If the session has gone badly, he
may elect to abort rather than log out. In that case his
current directory and any new files he has created are
discarded, and his starting directory and its files are
saved for the next session. In the event of system failure
his session is automatically aborted. In a long session
he may update his protection by logging out and back
in without incurring any special penalty.
THE EXECUTIVE AND
THE SPAT HARDWARE
As was stated at the outset, the planning for the
system involved continual interaction between the de-

135

signers of the software and of the hardware. This
section will consider those aspects of the hardwaresoftware system which, while elaborated in the environment of TX-2, may have application elsewhere.
These aspects have to do with the SPAT address-transformation hardware and its application in relocating,
bounding, and protecting the user's programs and in
facilitating the operations of the executive.
The SPAT transformation applies to all instruction,
deferred address, and operand references to memory.
It is effected in three stages. The first accomplishes the
switching of memory between the user and the executive. The second handles relocation, bounding, and
memory protection. The third gives memory paging.
In addition, because TX-2 is a mUltiple-sequence
computer (i.e., it has in effect many program counters) ,11 the transformation takes into consideration
which sequence (i.e., which "program counter") is
providing the location -for the current instruction. A
part of the computer called the Sequence Selector
handles the switching of program counters on demand according to a built-in priority rule. Most of the
program counters are tied to I/O devices or internal
sources of interrupts and have priorities based on the
speeds of the devices and the functions of the interrupts.
A few have no such associations and are used for
general computation. The flag bits that demand service for these latter sequences are entirely under program
control.
In the SPAT system, 29 of the 33 sequences are
grouped together and given the same transformation.
This grouping is indicated schematically in the box
labeled Sequence Selector in the upper left portion of
Fig. 4. When an instruction is executed in any of these
sequences all addresses are transformed according to
the information read out of the Executive Shelf in the
Location and Boundary Memory. Of the remaining four
sequences, the one called Startover is peculiar and is
not relevant to the executive system. It is not subjected
to the SPAT transformation. The other three sequences
are given individual Shelves in LBM and are treated
by the computer as user's sequences; i.e., privileged
instructions are prohibited. The present version of the
system uses only one of these three sequences; the
other two are treated as spares.
Since a sequence change in TX-2 is accomplished
in a time comparable to a single memory cycle, this
first stage of the SPAT transformation allows very fast
switching between user and executive. As will become
apparent presently, a Shelf in LBM corresponds to a
complete Map of core memory. So' the change of sequerices that occurs in response to an I/O event or an
interrupt automatically switches the address transform a-

136

A time-and memory-sharing executive program

SEQUENCE

l
I
A

APPARENT ADDRESS

SELECTOR

STARTOVER

...

ALARMS

1/0 INT'PTS

TRAPS

BOOK PAGE REGISTER
BYPASS SPAT TRANSFORMATION

r. J+-II

,....-----~

IlI
>-

0::

£r
(l.

DRUM

I *.
I
I
•
III
I
..HL _____
TYPEWRITER

EXEC I

LOCATION AND BOUNDARY
MEMORY (LBM)

....J

I
I

SHELF 0

SPARE (user) :I--t-....--It-+-.t:.

SPARE (user) :

I
I

SHELF

~_....-+_+-_T""'+-1""~
....

I
J~-

USER

EXEC

(exec)

SH ELF

(spore)

2 (spore)

TI....
'6\ORDS\ELF
SHELF 3

-+-+---~~~

(user)

-..
\01-....
\.+,1'1

LOCATION

r---~-..,

-I/O DEVICES (some spores)

, BOUND

r-..1-r--L--,

.I ADDER 'II1II........- - - - -....

FORBIDDEN BOOK

..--J

_1--.......

READ-ONLY BOOK ...

NON-EXECUTABLE BOOK ....--~

I-

'r-r-r'--.--.....

--.I

~ COMPARATOR J~+----------

BOUNDARY VIOLATION ...-...- - - - - - -......
WITHIN BOOK

::
_
-

t:::

PAGE
ADDRESS
MEMORY
'PAM)
1024 WORDS

STORE -TYPE INSTRUCTION

9
ACTUAL

Figure 4.

TX-2 address transformation hardware (SPAT).

,
8
ADDRESS

A time-and memory-sharing executive program
tion from the user's Map to the executive Map.
There are executive sequences with both higher and
lower priority than the user sequence. The sequences
of higher priority handle I/O and interrupts, and the
executive does its bookkeeping and scheduling in the
sequence labeled Exec I in Fig. 4. The lowest priority
sequence (Exec II) handles the user's calls to the
executive. A special timer measures the time spent
actually executing instructions in the User and Exec II
sequences, and this measurement is used instead of
real time to trigger a switch of users for time-sharing.
The second stage of the SPAT transformation is
governed by information that the executive has previously stored in the Location and Boundary Memory
(LBM). The information is automatically used to
perform relocation, bounding, and file protection during
every reference to core memory. The input to this
stage consists of the nine most significant bits of the
apparent address, which is illustrated in the upper right
corner of Fig. 4. The four most significant bits (the
book number) are used to select one of the 16 registers in the LBM shelf chosen by the first stage. The
five-bit page number in the apparent address is tested
against a five-bit boundary value stored in part of the
selected register of LBM, and a Boundary Violation
signal is generated if the address is outside the assigned
range for the book. There are also three control bits in
the LBM word. They can be set by the executive to
prevent the user's program from writing in the book,
executing an instruction in the book, or making any
reference whatever to registers in the book. This last
bit, which indicates a Forbidden Book, is used to indicate an unassigned book and to indicate a book that
has been assigned core pages but is waiting for data
to be transferred from the file memory. In the latter
cases, the executive will allow the user's program to
run in the absence of the file, but if his program refers to the file, the Forbidden Book bit is interpreted as
a signal to switch to another user if any others are
ready. The Read-Only control bit is ignored in the
Executive Shelf. This feature allows files like display
buffers to appear in both the user's Map and the
executive Map, appearing to the user as read-only but
appearing to the executive as writable.
Besides providing the control bits and the bound,
the second stage of SPAT performs the relocation that
provides segmentation into books. The selected register
of LBM contains a location (i.e., a base address) to
which SPAT adds the five-bit page number from the
apparent address.
Observe that this sum could have been used as the
final output of the whole transformation. With the last

137

eight bits of the apparent address appended to it, the
sum is a relocated address that has been checked for
boundary violations and could in principle be used
directly to address real memory.
But instead of being used directly, the sum of the
location and the apparent page number is used as the
input to the third stage of the transformation. It is
used to select one of the 1,024 registers of Page Address Memory (PAM). In that register the executive
has previously stored a 9-bit number, called the actual
page number. The last 8 bits of the apparent address
are appended to this number to form the 17-bit word
that is finally used to address the real core memory.
Conceptually, the third stage of the transformation
divides the real memory into pages, which are blocks of
256 consecutive registers. By storing the appropriate
information into PAM, the executive can define an arbitrary transformation between the page numbers that
enter the third stage and the page numbers that go
out to address the real memory. In particular, pages
that are in fact widely separated in real core can be
made to appear contiguous, simply by changing the
contents of the corresponding registers in PAM.
A word in Page Address Memory contains 12 bits,
3 more than are used to hold the actual page number.
Only one of the three is used in the present system. It
is set automatically whenever a store-type instruction
refers to a register in the page selected by the transformation. By inspecting these bits for all the pages of
a file, the executive program can determine whether the
contents of the file could have been altered. This feature can be used to avoid copying back onto the drum
a file that clearly has not been changed since it was
copied into core.
The main purpose of SPAT is to speed up the response of the system by (1) making it possible for the
executive to come closer to the ideal goal of keeping
the active part of each user's program in core and
ready to run, and (2) facilitating changes between
users and, especially, between user and executive. The
advantages of providing hardware for relocation and
bounding when more than one program is to be in
core are by now too familiar to require further comment. PAM was included to reduce the amount of time
that must be spent on storage allocation when the goal
is to keep as many programs as possible in core. Given
hardware for relocating 16 files independently, each at
its own base address, the problem of storage allocation
is primarily that of creating a block of contiguous
storage large enough to accommodate a file that must
be brought into core. With SPAT, the task of allocating
space in real core is replaced by the task of allocating

138

A time-and memory-sharing executive program

space in PAM. From the point of view of speed, there
is an obvious advantage in reducing by a factor of 256
the number of registers that have to be changed when
files must be moved to consolidate empty space into
larger blocks. There is an added advantage in the fact
that the number of registers in PAM is nearly three
times as large as the number of real pages in the TX-2
memory. This factor of three greatly increases the
probability that an adequate block of empty space will
be found without making any moves. Segmentation increases the probability further since it puts a limit of
32 on the number of contiguous PAM registers that
must be found to satisfy an allocation request. Assuming that PAM is already set up correctly, the change of
storage references required in switching from one user
to another is accomplished by storing into the User
Shelf of LBM the 16 words appropriate to the new
user's current Map. Switching between the user's Map
and the executive's Map is even quicker, since a
change of sequence automatically selects the proper
Shelf. Such rapid switching is highly desirable in a
system of this sort because of the large number of
I/O interrupts and executive calls to be processed.
The SPAT hardware could provide most of the control information needed for automatic page-turning,t7
but the auxiliary memory facilities in TX-2 are too
slow to justify this feature. While the hardware treats
a file as a collection of independent pages in core, and
the executive program handles it by pages in the file
memory, the executive does not now keep track of
the status of the individual pages which make up a
file. But given a suitable auxiliary memory, the system could be augmented without great difficulty to
include page-turning. In that case, the utilization of
the two spare User Shelves in LBM would provide the
user's program with nearly random access to almost
390,000 registers of apparent core memory.
CURRENT STATUS
At this writing, most of the executive program is
either operational or in the debugging stage. Parts which
are not yet ready for debugging are those having to de
with file groups and the handling of some of the shared
I/O equipment. Further work is also needed to replace the present primitive strategies for time-and
memory-sharing with more sophisticated ones.
ACKNOWLEDGMENTS
The system described in this paper has resulted from
the efforts of a large number of people both in programming and in hardware development. The original

design for the system was developed by a group made
up of D. B. Yntema, L. G. Roberts, J. E. K. Smith
(now at the University of Michigan), and the author.
The idea for the SPAT trnasformation was contributed
by J. M. Frankovich, and the design of it and the other
modifications to TX-2 were his responsibility. Much
of the detailed design of the executive program was
done by C. K. McElwain and A. N. Stowe. The program makes extensive use of the CORAL language
developed under the direction of L. G. Roberts. The
author is grateful to D. B. Yntema for his valuable
editorial assistance.
REFERENCES
1. W. A. CLAR, "The Lincoln TX-2 Computer
Development," Proc. Western loint Computer
Coni. (1957), pp. 143 - 45.
2. I. E. SUTHERLAND, "Sketchpad: A Man-Machine
Graphical Communication System," AFIPS Coni.
Proc (1963 Spring loint Computer Coni.) vol.
23, pp. 329 - 46.
3. T. E. JONSON, "Sketchpad III: A Computer
Program for Drawing in Three Dimensions," ibid,
pp. 347 - 53.
4. B. GOLD, "Computer Program for Pitch Extraction," I. Acoust. Soc. 01 Am., vol. 34, no. 7, pp.
916 - 21 (July 1962).
5. C. M. RADER, "A Speech Compression Simulation Compiler," presented at 69th Meeting of
Acoust.. Soc. of Am., Washington, D.C. (June
1965).
6. J. W. and C. D. FORGIE, "A Computer Program
for Recognizing the English Fricative Consonants
IfI and /0/." presented at 4th Int. Congo on
Acoustics, Copenhagen (Aug. 1962) .
7. L. D. EARNEST, "Machine Recognition of Cursive Writing," Inlormation Processing 1962, Proc.
01 IFIP Congress 62, North-Holland Publishing
Co., Amsterdam, 1963, pp. 462 - 66.
8. D. B. YNTEMA, "The Operation-Oriented Approach to On-Line Interaction between Scientists
and Computers," presented at meeting IEEE Prof.
Gp. on Human Factors in Electronics, Boston
(May 1965).
9. J. I. RAFFEL, et aI, "The FX-l Magnetic Film
Memory," Lincoln Laboratory Technical Report
#278 (Aug. 1962).
10. J. M. FRANOVIC and H. P. PETERSON, A Functional .Description of the Lincoln TX-2 Computer," Proc. Western loint Computer Coni.
(1957), pp. 146 - 55.

A time-and memory-sharing executive program

11. J. W. FORGIE, "The Lincoln TX-2 Input-Output
System," Proc. Western Joint Computer Coni.
(1957), pp. 156 - 60.
12. M. R. DAVIS and T. O. ELLIS, "The RAND
Tablet, A Man-Machine Communication Device,"
AFIPS Coni. Proc. (1964 Fall Joint Computer
Coni.), vol 28, pt. 1, pp. 325 - 32.
13. L. G. ROBERTS, "Machine Perception of ThreeDimensional Solids," Lincoln Laboratory Technical Report #315 (May 1963).
14. L. G. ROBERTS, "Graphical Communication and
Control Languages," Information Systems Scien-

139

ces: Proc. 01 the Second Congress, Spartan
Books, Washington, D.C. (to be published).
15. L. G. ROBERTS, "Graphical Communication in
a Time-Sharing Environment," presented at 1965
IFIPS Congress, New York (to be published).
16. T. E. JONSON, "Analog Display Generators,"
Lincoln Laboratory Technical Report #398 (to
be published).
17. T. KILBURN, et aI, "One-Level Storage System,"
IRE Trans. on Electronic Computers, vol. EC-II,
no. 2, pp. 223 - 35 (Apr. 1962).

Interactive machine-language programming*
BUTLER W. LAMPSON
University of California, Berkeley

INTRODUCTION

debugged in this way, and as time-sharing becomes
more widespread the interactive environment will become common.
It is clear that interactive debugging systems must
have abilities very different from those of off-line systems. Large volumes of output are intolerable, so that
dumps and traces are to be avoided at all costs. To
take the place of dumps, selective examination and
alteration of memory locations are provided. Traces
give way to breakpoints, which cause control to return
to the system at selected instructions. It is also essential
to escape from the switches-and-lights console debugging common on small machines without adequate
software. To this end, type-in and type-out of information must be symbolic rather than octal where this
is convenient. The goal, which can be very nearly
achieved, is to make the symbolic representation of an
instruction produced by the system identical to the
original symbolic written by the user. The emphasis
is on convenience to the user and rapidity of communication.
The combination of an assembler and a debugger
of this kind is a powerful one which can reduce by a
factor of perhaps five the time required to write and,
debug a machine language program. A full system
for interactive machine language programming (IMP),
however, can do much more and, if properly designed,
need not be more difficult to implement. The basic
ideas behind this system are these:

The problems of machine language programming, in
the broad sense of coding in which it is possible to
write each instruction out explicitly, have been curiously neglected in the literature. There are still many
problems which must be coded in the hardware language of the computer on which they are to run, either
because of stringent time and space requirements or
because no suitable higher level language is available.
It is a sad fact, however, that a large number of
these problem~ never run at all because of the inordinate amount of effort required to write and debug
machine language programs. On those that are undertaken in spite of this obstacle, a great deal of time is
wasted in struggles between programmer and computer
which might be avoided if the proper systems were
available. Some of the necessary components of these
systems, both hardware and software, have been developed and intensively used at a few installations. To
most programmers, however, they remain as unfamiliar
as other tools which are presented for the first time
below.
In the former category fall the most important features of a good assembler: 1.2 macro-instructions implemented by character substitution, conditional assembly
instructions, and reasonably free linking of independently assembled programs. The basic components of
a debugging system are also known but relatively unfamiliar. 3 • 4 For these the essential prerequisite is an
interactive environment, in which the power of the
computer is available at a console for long periods of
time. The batch processing mode in which large systems are operated today of course precludes interaction, but programs for small machines are normally

1. Complete integration of the assembler and the
debugging system, so that all input goes through
the same processor. Much redundant coding is
thus eliminated, together with one of two different languages serving the same purpose: to
specify instructions in symbolic form. This concept requires that code be assembled directly
into core-or into a core image on secondary

*This research was supported in part by the Advanced Research Projects Agency of the Department of Defense under
contract SD-185.

141

142

Interactive machine-language programming

-------------------------------------------------------

storage. Relocatable output and relocatable loaders are thereby done away with. (A remark on
terminology: It will be convenient in the sequel
to speak of the "assembler" and the "debugger"
in the IMP system. These terms should be understood in the light of the foregoing: different
parts of the same language are being referred
to, rather than distinct languages.)
2. Commands for editing the symbolic source program. The edit commands simultaneously modify
the binary program in core and the symbolic on
secondary storage. Corrections made during debugging are thus automatically incorporated into
the symbolic, and the labor of keeping the latter
current is almost eliminated.
3. A powerful string-handling capability in the assembler, which makes it quite easy to write
macros for compiling algebraic expressions, to
take a popular example, which can be handled
in a few other systems but rather clumsily. The
point is not that one wants to write such macros,
but that in particular applications one may want
macros of a similar degree of complexity.
These matters are discussed in more detail below. We
consider the assembler first and then the debugger
since the command language of the latter makes heavy
use of the assembler's features.
Before beginning the discussion it may be well to
describe briefly the machine on which this system is
implemented. It is a Scientific Data Systems 930, a 2microsecond, single-address computer with indirect
addressing and one index register. Our system includes
a drum which is large enough to hold for each user all
the symbolic for a program being debugged, together
with the system, a core image of the program and some
tables. Backup storage of at least this size is essential
for the editing features of the IMP system. The rest of
the system could be implemented after a fashion with
tapes.
THE BASIC ASSEMBLER
The input format of the assembler was originated on
the TX-O at M.LT. It has been adopted by DEC for
most of its machines, but is unknown or unpopular
elsewhere in the industry. Although it looks strange
at first, it has substantial advantages in terms of simplicity, both for the user and for the system. The latter
is a nonnegligible consideration, equally often ignored
and overemphasized.
The basic idea is that the assembler processes each
line of input as an expression (unless it is a directive

or macro call).5 The expression is evaluated and the
value is put into core at the word addressed by the
location counter. after which the location counter is
advanced by 1. Expressions are made up of operands,
which may be symbols, constants, numeric or alphanumeric and parenthesized subexpressions; and operators. Available operat()rs are +, -, *, /, .AND, .OR,
.NOT with their usual meaning and precedence; .E
(equals), .G (greater), .GE, .L, .LE, .NE, which are
binary operators with precedence less than +, and
yield 1 or 0 depending on whether the indicated relation holds between the operands or not; and #, a unary
operator with lowest precedence which causes its operand to be taken as a literal. This means that it is assigned a storage location, which is the same as the
location assigned to other literals with the same value,
and the address of this location is the value of the
literal. Blanks have the following significance: Any
string of blanks not at the beginning or end of an expression is taken as a single plus sign. An expression is
terminated by carriage return or semicolon. Several
instructions may therefore be written on one physical
line. This trivial feature proves in practice to have
significant advantages.
It is not immediately clear how instructions are conveniently written as expressions, and in fact the scheme
used depends on the fact that th~ object machine is a
single-address, word-oriented computer with a reasonable number of modifiers in a single instruction. It would
work on the PDP-6, but not on the IBM 7030.
The idea is simple: all operation code mnemonics are
predefined symbols with values equal to the octal encodings of the instructions. On the SDS 930, for instance, LDA (load A) is defined as 7600000 (all numbers are in octal). The expression LDA+200 then
evaluates to 7600200. When the convention about
spaces is invoked, the expression
LDA 200
evaluates to the same thing, which is just the instruction we expect from this symbolic line in a conventional
assembler.
Modifiers are handled in the same spirit. In the 24
bit word of the 930 there is an index bit, which is the
second from the left, and an indirect bit, which is the
tenth. With the predefined symbols
1==40000
X==20000000
the expression
LDA I 200 X
evaluates to 27640200. In more conventional form it
would look like this:
LDA* 200,2
There is little to choose between them for brevity or

Interactive machine-language programming

clarity. Note that the order of the terms in the expression
is arbitrary.
The greatest advantages of the uniform use of expressions accrue to the assembler, but the programmer
gains a good deal of flexibility. Examples will readily
occur to the reader.
Using this convention the implementation of the
basic assembler is very simple. Essentially all that is
required is an expression analyzer and evaluator, which
will not run to more than three or four hundred instructions on any machine. Because all assembly is into
core, there is no such thing as relocatability.
Two rather conventional methods are provided for
defining symbols. A symbol appearing at the beginning
of a line and followed by a comma is defined to be the
current value of the location counter. Such a symbol
may not be redefined. In addition, a line such as
SYM==4600
defines SYM. Any earlier definition is simply overridden.
The right side may of course be any expression which
can be evaluated.
The special symbol . refers to the location counter.
It may. ~ppear on the left of a == sign. Thus, the line
X;
.==. 40
is equivalent to
BSS 40
A
in a conventional assembler.
Note that the first punctuation character in a line of
input t

144

Interactive machine-language programming

In order to describe the result of using A after this
assignment, we introduce a distinction between the appearance of a symbol in a literal and in a normal context.
A symbol inside string brackets < > or single quotes
or in a macro argument is in a literal context; all other
contexts but one are normal. In a normal context, the
value of the symbol, whether· a string or a number, is
substituted for the symbol. In a literal context, on the
other hand, the characters of the symbol are passed on
unaltered. The case of a symbol on the left side of an
assignment is an exceptional one; such a symbol is of
course not normally ,evaluated.
To permit the value of a symbol to be obtained in a
literal context, the convention is introduced that a colon
preceding the symbol causes it to be evaluated if the
colon is at the top level of parentheses, brackets, and
quotes. If its value is a string, the characters of the
string replace the symbol; if it is a number the shortest
string of digits which can represent the number in the
prevailing radix replaces the symbol. Colon in a normal
context is illegal.
For convenience in delimiting string names a second
colon may follow a name preceded by a colon. This
second colon serves only to delimit the name and is
otherwise ignored. Thus if
AB ==
then
<:AB> == and <:AB:CO> ==
There are times when it is desirable to force evaluation of a symbol in a normal context when it would
normally pass unevaluated. The character & preceding
the symbol has this effect; it is exactly like : except
that it acts only in a normal context. Continuing the
previous example:
VW&AB == VWXYZ
and
&AB == 12
is equivalent to
XYZ == 12.
A string may be thought of as having two kinds of
structure:
1. It is composed of a sequence of characters.
2. It is composed of a sequence of substrings delimited by commas not enclosed in parentheses,
brackets, or quotes.
With reference to the first structure, a single character
may be selected by a subscript enclosed in brackets.
Referring to the string assigned to A, we note that
A[2] is <,>, A[6] is , and A[7] is <».
By an obvious extension of this notation,
A[3,7] is «C,O» and A[9,11] is .
Subscripts which reference the substring structure
are enclosed in parentheses. Thus
A(I) == and A(2)== .
Note that a single pair of parentheses surrounding a sub-

string is removed. Subscripting may be iterated:
A(2) (2) == <0>.
Subscripting is applied only to a string-valued symbol
which is in a normal context or is evaluated by a colon.
Subscripting of a name on the left side of an assignment
forces it to be evaluated even if it is not preceded by
a colon.
Two operations, .L and .LC, determine respectively
the number of substrings and the number of characters
in their arguments. Thus
.L(A) == 4, .L(A(2)) == 2 and .LC(A) == 11.
Having dealt with the general machinery for handling
strings, we now turn to the slight refinement which adds
macros with arguments to the system. This takes the
form of a modification to the ordinary line assigning a
string to a symbol, which permits an argument string to
be specified. Thus
STORE ==
<.RPT.FOR T == 1, .L(ARG(2) ),1

defines a macro with two arguments, the first a string
which, when appended to , creates a store instruction, and the second a list of 10catiQJ).s to be stored
into. Whenever STORE is used, the string of characters beginning with the first following nonblank character and ending with a line delimiter or unmatched
right parenthesis is made the value of ARG. The string
which is the value of STORE is then substituted for it
as usual.
STORE might be called with
STORE A,(Sl,S2,S3)
which is, because of the definition, equivalent to
.RPT.FOR T == 1,3,1
(T»
To complete the expansion we must consider the
.RPT directive which has been, used above. This directive causes the string which follows to be scanned
repeatedly .. It takes one of two forms:
1. .RPT N < ... >
which causes N repetitions, or
2. .RPT.FOR J == nl,n2,n3 < ... >
which causes (n2-n1) /n3 + 1 repetifions with J initially set to nl, and then incremented by n3 until it exceeds n2. Zero repetitions are possible. The n3 may be
elided if it is 1.
The STORE macro call above may' now be seen to
expand into
STA SI
STA S2
STA S3
We illustrate with two further examples. The first is
a generalized MOVE macro which takes as its arguments a sequence of pairs of lists. The first list of each
~

Interactive machine-language programming

pair specifies the locations to load from, while the second gives the corresponding locations to store into. A
list may of course have only one element.
MOVE =
<.RPT.FOR SI = 1, .L(ARG),2
*THIS LINE STEPS THROUGH THE PAIRS OF
LISTS
<.RPT.FOR S2 = 1, .L(ARG(SI»
*THIS LINE STEPS THROUGH THE ELEMENTS
OF ONE PAIR OF LISTS
< LDA ARG(SI)(S2)
< STA ARG(SI + 1) (S2) >:»
thus
MOVE A,B,C,D
becomes
LDA A
STA B
LDA C
STA D
So does
MOVE (A,C),(B,D)
Suppose that we have some two-word data structures
to manipulate. We can attach to the name of each structure a string of the form . A is the address of
the first word of the structure, B of the second. A
macro can do this and assign the storage.
TW =
<
TWSI = TWS + 1
ARG(I) =
TW&TWS,O
TW&TWSl,O
TWS=TWS+2 >
Now, if we call TW twice after setting TWS to 1:
TW
A
TW
B
we will have given A the value and B
the value and defined the four TW symbols.
We can now use A and B in the MOVE macro. In
fact
MOVE A,B
expands to
LDA TWI
STA TW3
LDA TW2
STA TW4
With the addition of one more device we can proceed
to the definition of a very grandiose macro. The directives .IF and .ELSF, used thus:
.IF
El < ... >
.ELSF E2 < ... >

145

.ELSF En < ... >
cause each El in tum to be evaluated until one is greater
than zero. The string following this one is then scanned
and the rest of the structure ignored.
*THIS MACRO .C -

< EXPR-<:ARG(l).>

*APPIIID • TO THE EXPRESSION
*INITIALIZI THE STACK WHICH lWIDLES
*PAUllTHESIS
*IIIITIALIZI THE CIWIACTIR POlllTER
*IIIITIALIZI THE TIIIl'ORARY STORAGE COUNTER

STK-<*>
Jol

n-o

*IF TEMPORAllY STORAGE IS RE~IIBD IT IS ASSIGllED AS TEMPl,
*TEMP2, ETC., AND TI KEEPS TRACK OF THE IIEXT AVAILAILE LOCATION.
*THIS IS THE MACRO WHICH DOES THE WORK
4RROR> >

Xl
.IF T .IIE '.'

*CHECl( THAT EXPRESS 1011 WAS 1I0T TERMlllATED BY A RIGHT PAIBIITHESIS.
*THIS MACRO COLLECTS A SUB-EXPRESSION COIISISTIIIG OF OPERANDS
*STRUNG TOGETHER WITH + AND -. IF THE SUBEXPUSSIOII IS A SIIIGLE
*VAIlIABLE, COP (CURREIIT OPERAND) WILL BE THAT VAIlIABLE 011 EXIT.
*OTHERWISE IT WILL BE EMPTY.
<

Xl COP - <***>

*EIISUIB THAT COP IS 1I0T EMPTY IIIITlALLY

*AN EMPTY COP MEANS THAT CODE MAS BEEN ASSEMBLED LEAVING A VALUE
*IN n!E A REGISTER. IF COP IS A LETtER, IT IS THE VAIlIABLE
*WHICH IS THE CURREIIT OPERAND.
OPERAND
.RPT .FOR E-l,l,O
<

*GET THE FIRST OPERAND
*E IS SET TO 2 WHEN THERE AIlE NO MORE
*SIGNS
*EXPECTIIiG AN OPERATOR OR TERMINATION

To':EXPR[J:'
JoJ+l
.IF T .E ' . ' .01. T .E ')'

OR -

*SET E TO TERMINATE THE LOOP IN THIS CASE.
.ELSF
.ELSF

T.E '+'
T.E '-'

*IF A + OR - IS PRESENT, GET THE SECOND OPERAND AND COMP lLE CODE.
.!LSF 1
> >

*OTHERWISE, ERROR
*CLOSE LOOP AND MACRO

*THIS MACRO COLLECTS THE SECOND OPERAND OF A BINARY OPERATOR AND
*CONSTRUCTS CODE TO PERFORM THE SPECIFIED OPERATION. IT USES ITS
*FIRST AIlGUMENT IF THE FIRST OPERAND IS IN THE A REGISTER, ITS
*SECOND AIlGUMENT IF THE SECOND OPERAND MUST BE IN A AND THE FIRST
*TAl OPERAND
.IF .LC(COP) .G 0

*GET THE SECOND OPERAND

*111 TIIIS CASE THE SECORD OPD.AIIID IS A SIIIGU VAIllAl1.l.

•IF

.LC (PUVOP)

.G 0

l'UVOl'>

*IF TIl! FIRST OPERAND IS ALSO A VAIlWLI (Ia A 'IIIIr LOCATIOlI)
*BRIIIG IT IIITO A
CAIG(l)
.ILSF 1

COP >
*AND CGIl'ILI COOl

*OTHEIlWISE THE SECOII) OPERAND MIIST II III A, AND THE PWT III MEHOaY

GOP...: >

*SET COP TO IIIDlCATI A VALUE III A AND CLOSI THE MACRO.
*THIS MACRO COLLICTS All OPIIARD, WHICH MAY U
*SUBI!U'RISS 1011

A PAREIITHESIZID

OPERAND<

T - ':UPR[J)'
J-J+ 1
.IF T .E '('
.IF .LC(COP) .E 0

*GET THE IIEXT CIWIACTD
*IT SHOULD BE A LlTrU 01 (

*IF WE AJ.RIWJY HAVE A VALlIE III A IT MUST BE SAVED III TEMPOIAIlY
*STORAGE WHILI THE SUBUPlBSSIOII IS EVALUATED.

< TI - TI +1:
*CONSTRUCT A TEMP LOCATIOII TO SAVE IT III
*AlID ltEMEMBER IT III COP
*STICK COP 011 THE FROlIT or sm

STA TEMP"TI
COP- >
STR-<:COP,:STK>

Xl
.IF T .IIE '('
1-1
*RESET THE TERMlllATIOli SWITCH FOJ. Xl
PREVOP-<:STK(l)>
*SET PREVOP TO THE OLD COP WHICH WAS SAVED
STK-<:STK(2, .L(STK»> >

iOUKOVE OLD COP FRIM STK AND 'l'E\UIlHA'l'E THIS CASE.
.ELSF

T .GE 'A' .AlID T .LE

Xl MAS SET COP

'z'

*IF T IS A LETTER (RECALL THAT THE CIWIACTIIl CODE IS ASCII)
<

PREVOP-<: COP>
COP-<:EXPR[J-l»
.ELSF 1 qRROR>

>
>

146

Interactive machine-language programming

This macro, called by
ARITH «A + B) - (C- D»
would generate
LDA A
ADD B
STA TEMPt
LDA C
SUB D
CNA
ADD TEMPt
Note that there are only three lines in the definition
which actually generate code. The temporary storage
location TEMPt must be defined elsewhere.
The implementation of all this is quite straightforward. When a string is encountered, it is collected
character by character, due attention being paid to
colons, ampersands, brackets, and quotes, and stored
away. When it is referenced, the routine which delivers
characters to the assembler, which we will call CHAR,
is switched from the input medium to the saved string.
This process is of course recursive. When the string
which is the current source of characters ends, CHAR
is switched back to the string it was working on before.
All the various occurrences of strings are treated perfectly uniformly, except that in the case. of macro
definitions the substrings of the argument string are
delimited when the latter is collected to improve the
efficiency. Perfectly arbitrary nesting of the various constructs is possible because of the recursiveness of the
string collection and reference routines.
In the interests of efficiency the .IF directive is not
handled in this way, since its subject string is scanned
either once or not at all. All that is necessary is a flag
which indicates whether an .ELSF directive is to be
considered .or ignored.
THE DEBUGGING SYSTEM
An interactive debugging system should not be designed for the occasional user. Its emphasis must be on
completeness, convenience, and conciseness, not on
highly mnemonic commands and self-explanatory output. The basic capabilities required are quite simple in
the main, but the form is all important because each
command will be given so many times.
One essential" completely symbolic input and output
is half taken care of by the assembler. The other half
is easier than it might seem: given a word to be printed
in symbolic form, the symbol table is scanned for an
exact match on the opcode bits. If no match is found,
the word is printed as a number. Otherwise the opcode
mnemonic is printed, indirect and index bits are checked,
the proper symbols printed, and the table is scanned for

the largest symbol not greater than the remainder of
the word. This symbol is printed out, followed if necessary by a + and a constant.
The most fundamental commands :are single characters, possibly preceded by modifiers. Thus to examine
a register the user types
/xt-3; LDA I NUTS + 2
where the system's response is printed in capitals. This
command may be preceded by any combination of
modifiers:
C for printout in constant form
S for printout in symbolic form
o for octal radix
D for decimal radix
R for relative (symbolic) address
A for absolute address
H for printout as ASCII characters
I for printout as signed integer
N for no printing of addresses
L (load) for no printing of register contents
The modifiers hold until the user types a carriage return or gives another / command.
For examining a sequence of registers, the commands
+ and - are available. The former examines the precediJ:~g register, the latter the following register. In the
absence of a carriage return the modifiers of the last
examination hold. The ~ command examines the register addressed by the one last examined.
The contents of a register may be modified after
examination simply by typing the desired new contents.
Note that the assembler is always part of the command processor, and that debugging commands are differentiated by their format from words to be assembled
(as noted above, an assembler line has comma or space
at its first punctuation character, and all debugger lines
have some other initial punctuation character). Furthermore, debugging commands may occur in macros, so
that very elaborate operations can be constructed and
then called on with the two or three characters of a
macro name.
To increase the flexibility of debugging macros, the
unary operator @ is defined. The value of @ SYM3
is the contents of location SYM3. With this operator,
macros may be defined to type out words depending on
very complicated conditions. A simple example is -

TG
.RPT.FOR TEMP

= A(1),37777,1

*SCAN THROUGH ALL OF STORAGE STARTING
AT THE LOCATION GIVEN BY
*THE FIRST ARGUMENT

.IF @ TEMP E. (A)2
*IF THE CURRENT LOCATION MATCHES THE

Interactive machine-language programming
SECOND ARGUMENT, THE SCAN IS OVER
1

--16

Instruction Formats:

7,8-11,
X2
-OP
12 ,

'-I

- 17,18

32-35

Xl
II
II

ADDRESS

II
II

ADDRESS
ADDRESS

XO
10

Scalar RegisterRegister
Scalar RegisterMemory
Vector
Transfer

0-3;4-7

Control Words:

15,16-17,18
35
C
Instruction
C
Counter
Interrupt
Condition Mask
Branch Adr.

Interrupt
Conditions

(CC - condition code)

Program Status Word
Program Status Word
Mask

Errata
line 10 of Fig. 6: Change to B ( 1) to ELEMENTS
line 11 of Fig. 6: Change to B(A2(1», B(A2
(I + 1»,
line 12 of Fig. 6: Change to A2 + 1,2,,1
1),A2(I 2), ... ,
ADD A2(1
A2(1 +- 16)
line 13 of Fig. 6: Change to A2-1,2,,1 ADD
A2(1- 1),A2(1), ... ,A2(1 + 14)
line 15 of Fig. 6: Change to A2(1
1),... ,
A2(1 + 15)
.

F. J. CORBATO and V. A. VYSSOTSKY
Introduction and Overview of Multics System
p. 186, line 6, R: Change Refs. 23 and 24 to Refs. 42
and 24
p. 190, line 45, L: Change more compilations. to
more compilation time.
p. 196: Insert at bottom, Ref. 42 as: A. L. Samuel,
"Time-Sharing on a Computer," New Scientist 26,
445 (May 27, 1965) 583-87.

E. A. BOWLES
The Role of the Computer
p. 270, line 51, L, last word: Change oppositions to
opposition
line 10, R: Change kowing to knowing
line 15, R: Change Blender to Blendor
line 20, R: Delete Old
p. 272, lines 38 and 39, L: Reverse their order
line 6, R: Insert comma after tables
p. 273, line 18, R: Change separate to different
p. 276, line 3, R: Change Pres to Press
line 11, R: Italicize and capitalize Ibid
D. C. CLARKE and R. E. WALL
An Economical Program •••
p. 313, footnote to Table 1: Should be associated with
Table 4, p. 314.
ZWICKY et al
MITRE Syntactic Analysis Procedure
p. 320, lines 43-47,R and p. 321, lines 1-4, L: Correct format to read as follows:
Phrase Structure Component:
75 rules

167

approximately 275 subrules
Transformational Component:
13 initial singularies
26 embeddings and related
singularies, including
9 embeddings
15 final singularies
54 rules
p. 321, line 34, L: Change placement of OPEN to
read
PRES sa VTR
OpkN
p. 322, line 17-19, L: Add lines to diagrams to read
(A) ADRIS n
(1 B C) SUB n

/\1

231

p. 322, line 26, R: Correct "Checking by Synthesis"
as secondary heading (as in line 18), italized, flush
left, and with following text entered with paragraph
indentation.
p. 323, line after line 12, R: Insert "OTHER APPROACHES TO THE ANALYSIS PROBLEM" as
major heading (as in line 26, L above-"Areas for
Further Investigation") above "Analysis by Synthesis"
p. 325, lines 12-29, L: Correct format to read:
(Forward) Grammar
Phrase Structure Component:
61 rules
105 subrules
Transformational Component:
11 initial singularies
6 embeddings and related
singularies, including 2
embeddings
3 final singularies
20 rules
Surface Grammar
32 rules
306 subrules
Reversal Rules
6 final singularies
15 embeddings and related
singularies
11 initial singularies
32 rules
R. C. MINNICK
Cobweb Cellular Arrays
p. 328, line 8, L: Change index to Index
p. 329, caption for Fig. 2: Change one to One
p. 329, caption for Fig. 3: Should read Cutpoint Realization for a Three-Bit Parallel Adder
p. 330, caption for Fig. 4: Should read Cutpoint Reali-

168

p. 331,
p. 332,

p. 333,
p. 334,
p. 335,

p. 336,

p. 337,

p. 338,
p. 339,

p. 340,

Errata
zation for a Five-Bit Shift Register
caption for Fig. 5: Should read Cutpoint Realization for Three Functions of Three VariabIes
caption for Fig. 6: Should read Structure of the
Cobweb Array
caption for Fig. 7: Should read Diode-Transistor Realization of Cobweb Cell
line 17, L: Poorly printed word is cutpoint
Eq. (3): Replace (j with EB
line 1, R: Replace (j with EB
caption for Fig. 8: Should/read Shannon and
Reed Decompositions Using Cobweb Arrays
caption for Fig. 9: Should read Cobweb
Realization for a Three-bit Parallel Adder
caption for Fig. 10: Should read Cobweb
Realization for a Five-Bit Shift Register
line 21, R: Change focal to local
caption for Fig. 11 : Should read Cobweb
Realization for Three Functions of Three
Variables
caption for Fig. 12: Should read Cobweb Supercells
caption for Fig. 13: Should read Exhaustive
Listing of the Cobweb Array Fault-Avoidance Algorithm
caption for Fig. 14: Should Read Block Diagram for a Twelve-Bit ·Serial Multiplier
caption for Fig. 15: Should read Reali~ation
of the Multiplier in Terms of Five Cutpoint
Arrays
caption for Fig. 16: Should read Realization
of the Multiplier in Terms of One Cobweb
Array

R. H. CANADAY
Two-Dimensional Iterative Logic
line 23, R: Change to AB + Ac + Be
line 17, R: Change to r(xl, ... ,xn) == f(xl, ... ,Xn)
line 17, L: Change Xf/ to Yf/
line 25, R: Change C to Cline 26, R: Change C to C and U to U
line 27, R: Change U to U
line 33, R: Change to f Sd == BCDU + ABCnU
+ ABCD + ABCD + ABC
line 34, R: Change to + ABeU +ABCU
+ ABCDU + BDU + COU
.
p. 349, line 11, L: Change {B,B,U,U} to {B,B,U,U}
line 31, L: Change {C,C,O,I} to {C,C,O,I}
line 7, R: Change (n-l to (n-l)
line 22, R: Change to gt == BCgtoo + BCgtot +
BCgll0 + BCg111

p.
p.
p.
p.

343,
344,
345,
346,

Array 5, R: Change element 1,6 from B to
C
p. 350, Array 7, R: Change element 2,6 from gU01
to g1100
line 15, R: Change A + B to A EB B
line 16, R: Change + C +D to EB C EB D
p. 351, line 4, L: Change to {xt,x1x2,x2, ... ,x(n-l),U,U)
(U,U) being
Array left boundary, L: Change all 'V to U
p. 352, line 17, L: Change UU to UU
line 18, L: Change UgOd to UgOd
line 19, L: Change UU to UU
line 20, L: Change Ug1d to U g /
line 5, R: Change UgOd to UgOd
line 6, R: Change U g / to U g /
line 10, R: Change UU to UU
line- 22, R: Change to gosd == URogoo + USog01
+ URogd01 + USot00 + URoSo
line 24, R: Change to glsd == URtg10 + UStgll
+ URlglld + USlglOd + URlS1
line 29, R: Add the term VXRoSo
line 30, R: Change to + VYS1gU + VYRtS1
+ WXRltll + WXSlGd lO
line 31, R: Change gOl to tOl and goo to too
line 33, R: Change UU to DU
line 35, R: Change UU to Vu; following that
line should be the heading SUMMARY
line 41, R: Change AB+AC+BC to AB+
AC+BC
p. 353, line 1: Eliminate SUMMARY

R. A. SHORT
Two-Rail Cellular Cascades

p. 360, line 12, R: Change f to
line 15, R: Change f to f

p. 361, lines 4, 5, 6 L, eq. for f (x!)'''' x n ) : Change all
+ to EB
line 10, L: Change f2: (Y2, Y2 + XY1) to f2:
(Y2, Y2 EB xYt)
line 20, L, eq. for N: Change (+) to e)
line 7 from bottom, R: Change + to EB
p. 362, line 3, L: Change + to EB
line 5, 6, 7 from bottom, L: Change all
EB
line 2, R: Change e~3) to ('1)

+ to

p. 363, line 2, 3 from bottom, L: Change all + to EB
p. 365, line 5, 6 from bottom, R: Change all
p. 367, Fig. 11 (f): Change gz to Y2

to EB

Errata
G.E.ROUDABUSHd~

Left Hand of Scholarship • • •

p. 399, line 2, title: Change MEDIA to MEDIUM
p. 401, line 39, L: Omit the
line 33, R: Space between we and received
p. 402, line 2, L: Change PC to PC6
p. 403, line 6, L: Change printd to printed
p. 404, line 16, R: Change gold to bold
p. 405, line 2, L: Change slit to split
p. 406, line 20, L: Change in our office to done
line 11, R: Omit that
line 12, R: Change dattum to datum
p. 408, line 13, L: Change see to set

D. ROOS
An Integrated Computer System • • •

p. 424, Fig. 1, R: Distance of 500.26 left out in figure
p. 426, line 50, L: Change sybsystems to subsystems
p. 427, line 38, R: Change DO I L == 1, I to DO 1 L
== 1, I
p. 428, line 6, R: Change DOP to DO
p. 430, line 38, R: Change 'N POINT' to 'NPOINT'
p. 432, line 8, L: Change System 360 to System/360
line 11, L: Change System 360 to System/360
p. 433, line 5, R: Replace line in 4 with name R. D.
Logcher

C. W. ADAMS
Responsive Time-Shared Computing • • •

169

line 19, L: Change T to 1:
line 23, L: Change to g == 3T - T
line 24, L: Should read As T is made to approach I, G will approach Q,
line 25, L: Change G to G
line 26, L: Change to G. == 2T
line 42, L: Change 4T to 4T
line 7, R: Change 4T to 4T
line 25, R: Change our to out
line 27, R: Change (Xl) to (Xl)
p. 497, line 6, L: Change 4T to 4!; and where T to
where I
line 12, L: Should read alter T and T
line 10, R: In two places change 4T to fi
line 12, R: Should read 3T - I
line 18, R: In both places change to 3T - T
line 22, R: Should be 3T - I
Change 4T to 4I
line 23, R: Change TIT to TIT
line 24, R: Change T to T
line 27, R: Change 4T to 4T
line 28, R: Change T to T
line 30, R: Change T to I; and TIT to TIT
line 32, R: Change T == to T ==; and 3T to 3T
line 33, R: Change TIT to TIT
line 39, R: Change T to T
line 43, R: Change T to I.
line 45, R: Change T to I
p. 498,. line 42, L: Should read clock system of w.. ==
line 4, R: Change to (Tl + T2 + Ta + T.)
line 7, R: Change (4T) to (4I)
line 11, R: Should read .I. == 2.0
line 17, R: Change to (T1
T2
Ta) - I
line 23, R: Should read has negligible T - T
line 26, R: Should be (T2 + Ta)
line 27, R: Should read with negligible T - I
p. 499, line 4, L: Should read W2 == 4I. - W1 == 41
- We - Se
line 9, L: Should read of 4 I. + Z; and delay
is Sc + 4T + Z
.
line 10, L: Should be Wa == (Sc + 4T + Z)
- (4I.+ Z)
line 11, L: Should be == 4(T - I) + Sc
line 13, L: Change W4 to WI
line 15, L: Should be

+ +

p. 486, line 6, R: Should follow line 8

L. W. COTTEN
Circuit Implementation • • •
p. 489, line 19, L: Change 23 to 20
p. 491, line 32, L:
n-k

P12 == l AtBi + k
1=1
p. 492, line 36, L: Change NOR-OA to NOR-OR
line 27, R: Change f == AB to f == Ali
line 28, R: Change + AB to + AB
p. 494, line 2, L: Change (X) to (X)
line 36, L: Change (Xl -) to (Xl -)
line 34, R: Change G to G
line 35, R: Change to G: == C + D + A
line 41, R: Change to G: == H - H + I + E
p. 495, line 5, L: Change to Q == K - K + M + T
line 18, L: Change T to T

.c.

- (We +2Sc)+ (4!-Wc- Sc) +4(T-I)
line 18, L: Should read When 41
line 22, L: Change to f == II (Sc + 41')
line 23, L; Should read When 4I
line 25, L: Change term (T - T) to (T - I)
line 27, L: Change If 4T to If 41

170

Errata

Change then T == to then I ==
line 28, L: Should read If this I. is
line 30, L: Change both 4T terms to 4T
line 31, L: Should be It is seen that I
line 33, L: Should be When 4I
line 36, L: Change 4T to 4:[
line 2, R: Change both 4T terms to 4I
line 4, R: Change 4T term to 4T; Should read
speed T might
line 6, R: Should read If T == 4.5, I. == 2.5,
line 10, R: Change term 4T to 4T
line 11, R: Change term in parens to (T - I)
line 15, R: Should read such that the r.
p. 503, line 26, R: Change Relay to Delay
line 29, R: Change T to T
line 31, R: Change T to T
line 35, R: Should read X: Logic sense
line 38, R: Should read X: Minimum value

G. H. BALL
Data Analysis in the Social Sciences • • •
p. 533: Delete entire left column, replacing the two
existing paragraphs with the three following
paragraphs:
The advent of the relatively inexpensive digital computer
makes iterative cluster-seeking methods of analyzing complex multivariate data practical when they could not be
seriously considered before. These "cluster-seeking" techniques provide a way of viewing multivariate data that
differ from factor analysis and discriminant analysis.
Cluster-seeking techniques are best suited to examining
problems where the data are multi-model. They provide a way
of detecting isolated data points that are not "close" (relative
to the data set) to any other points. These techniques can
be used to show the relationship of a single data point to
the entire set of data-thus allowing an examination of the
details in the data. Large numbers of data points can be
structured and related to each other in the original highdimensional space.
We feel that the techniques described below provide a
useful adjunct to other methods of analyzing multivariate
data. We compare and describe below, a number of these
methods reported in the literature.

p. 534, L: A portion of Fig. 1 has been omitted. See
the accompanying figures for complete Fig. 1.
R: Delete last paragraph and replace with the
following text:
A useful description of all the data can be plotted using
the axes found by factor- analysis or principal components
analysis. Even this may not be as meaningful as desired as
can be seen by examining Figs. 2a and 2b where we see that
two quite different two-dimensional probability distributions
both give rise to marginal distributions that are uniform
along vertical and horizontal axes. We can see that this inability to distinguish between these different distributions

can be resolved by using more axes to describe the data (as
shown by the marginal distributions along the diagonal axes).
Relating events on the different axes, particularly when the
data are high-dimensional is difficult, however. The major
problem, then, is that two sets of data that are. quite dissimilar can appear to be quite similar when viewed by data
analysis techniques implicitly oriented toward Gaussian distributions.

p. 535, L: Delete lines 1, 2, and 3.
R: A portion of Fig. 2b is missing in print.
See the accompanying figures for complete
Fig. 2.
p. 538, L: Delete the second paragraph and replace
it with the following:
The sample size requirement relates to any quantities that
must be estimated (in a statistical sense) from the set of
sample patterns. Allais3 showed that this requirement cannot be taken lightly. He shows that given N samples the
estimation of a covariance matrix of dimension greater than
N /10 usually increases the probability of error for predictions based on that covariance matrix as compared with
predictions using a covariance matrix of fewer dimensions.
Small sample sizes seem to require that simpler quantities
be estimated-such as the means of clusters, or only the
largest eigenvector of the covariance matrix.

p. 539, Table 2 (section Weighted Boolena "and":
Add a dot to show multiplication of terms.
I,Xkj • Xl j == 0
(Section "Normalized correlation"): Equation should read I == 1
p. 540, Table 3 (section "Entropy"): Change equation i == 1 to j == 1
(Section "Square (used for deviation from
single signal)"): Equation on last line
should read j == 1
(Section "Value for a cluster"): Equation
starting with Ix! change to Ixy
(Section "Coefficient of belongingness"):
Term shown on last line should be i,j~Ck
p. 541, table (section "Total entropy"): term
- 1 I
-1 I'
..
. should be
-2··
2 q
q
In the other two instances where the symbol
I is used, prime symbol should be at right,
thus: I'
p. 543, line 20, L: Change Mattson (1965) to Mattson & Dammann (1965)
p. 546, line 8, L: Add stored in memory so that the
line will read as follows: fraction of the
waveform stored in memory as it exists at
the time the
p. 547, lines 44 and 45, L: Change troubles to
troubled
p. 548, L: Change the following lines to read:
line 20: cluster is broken into two clusters by
creating two

Errata
p: 1/4

1.0

P:1/4

2.0

-2.0

171

P:1/4

-1.0

I [-r4 oi~']

0.6~:[·Ke-Z xlo 1Jx
'A

-2.0

~--

'+1.732

. .L

-1.97

-2'

-I'

1.97

\1
"2

UNIFORM
"- ~ PROBABIliTY
DENSITY. TOTAL
PROBABILITY OIF 1/2

Fig. 1

---

- - OF DATA-HAVING
-THREE-DISTRIBUTIONS
- THE SAME
COVARIANCE MATRIX

1-1•112

UNIFORM
PROBABILITY
DENSITY. TOTAL
PROBABILITY
OF 1/2

172

Errata

THIS NUMBER IS
THE PROBABILITY OF A
DATA POINT LYING IN
THIS SQUARE. THE POINTS
ARE UNIFORMLY DISTRIBUTED
WITHIN EACH BOX .

• x,
Fig. 2b

A UNIFORM, TWO DIMENSIONAL DISTRIBUTION OF POINTS
HAVING UNIFORM MARGINAL DISTRIBUTION

line 21: cluster points out of the original cluster. A second
line 22: allowable number of patterns in a
cluster, the allowable maximum
p. 550, line 17, L: Change to read: around this by
utilizing an arbitrary threshold. All
p. 552, lines 1-8, L: Change to read:

normally distributed so that correlations have the desired
significance. It seems quite probable that in cases in which
a relatively large number of patterns is being used and in
which a priori knowledge is not really adequate to divide
these patterns into homogeneous subsets that this will not be
true.

line 17, L: Correct spelling of phenomena
line 49, R: Change to read tween the technique
of Mattson & Dammann27 and that of Cooper

Errata
line 51, R: Change to read son and Dammann
have taken the underlying philosophy of the
p. 553, line 2, L: Should read the Mattson and Dammann technique particularly useful. The
most
p. 554, line 38, iR: Alphnumeric should be alphanumeric
p. 555, L: Change the following lines to
line 12: distance of all patterns from this average. Set a
line 14: from this overall average) where
o L. k. All
line 18: allowable cluster size is 1. The overall
average of
line 19 L: Add all patterns would be used as
the initial cluster point.
p. 558, line 11, L: Add n to name Dammann.
A. R. HOLMES and R. K. AUSMAN
Information Processing of Cancer Chemotherapy Data
p. 583, line 4, L: Change chemotheraphy to chemotherapy
p. 585, line 51, R: Change fransit to transit
p. 586, line 12, L: Change deliquent to delinquent
line 1, R: Change Institutes or to Institutes of
line 15, R: Change mM1620 to mM 1620

N.V.FINDLER
Human Decision Making •••
p. 738, line 25 and 26, R: Change to Logic Theorist
or the General Problem Solver
p. 739, line 19, R: Change to {y} == F( {x})
line 40, R: Change to to this, the concept
p. 741, line 40, R: Change behaps to perhaps
p. 743, line 6, L: Change ding on sich to Ding on sich
line 5, R: Change etc.; Hooke to etc., Hooke
last line, R: Foot,!;ote omitted:
*One subject, disregarding the fact that there were three
independent variables instead of two, considered the environment as some terrain and the £ values representing
the height of mountains, hills, and valleys. His "hillclimbing" was a tourist's excursion.

p. 744, line 32, L: Change the one to the ones
last lines, L : Footnote not needed
line 29, R: Indent to start under "Well
p. 746, line 17, L: Change lowing to lowering
p. 748, last line, L, and first line, R: Write last line
of formula as:
8.sin k 1[. 10. 8k ,o(mod 1),

line 7, R: Change to
Y2 == 250 - 200. 181

p. 804, line 1, R: Should follow line 12

> 2-

8==0

new line 44, L: Change to 1
line 24, R: Change 4 to 5
line 25, R: Change 5 to 6

p. 725, line 3, L: Change 2. _ _ to 2. M. Lehman

D. E. RIPPY and D. E. HUMPHRIES
MAGIC-A Machine for Automatic Graphics •••
p. 820, lines 35-37, R: Change (0 The Z Field specifies
(Continued page 180)

APPENDIX

III

Comparison between Excerpts from a Representative Protocol and Computer Performance
In both instances, in the experiment and in its computer simula-

tion, Task Environment 1 was used with the simplest cost function, in which
c;= Yl.

Editorial remarks are put in square brackets.

The subject's initial search here could be classi-

fied as systematic with slight random components.

-;~

He adopted the Mathematical Language of Represen-

tation without any hesitation.

In describing the computer's actions, instead of articulated English sentences, we can only refer to brief

In Stage (A) he follow~ statements, such as EXTRAPOLATION, INTERPOLA-

ed a scheme fairly thoroughly but was not satisfied

with the rate of improvement and entered Stage (B).

"" NEW POINT:, etc.

In this, he employed the so-called univariate method

* gram to indicate at appropriate times what action the ma* chine is taking. The following excerpts from the trace of
* the program (p) are approximately equivalent to the seg* ments of human behavior at the left, almost paragraph by
* paragraph.

which consists of changing only one control variable
at a time.

The optimization takes place in terms of

single control variables and at the end an attempt is
made to specify a global optimum.

* * *
E:

Can we start now?

O.K., let's try around zero, say, 2, 3 and 4.

TION, ASPIRATION LEVEL NOW:, COUNTER NOW:,

*
*

These had been planted in the pro-

* * *

INITIAL SEARCH TYPE 1.

NEW POINT:
OUTPUT:

Xl= 3, X2= 47,
Cl=273, C2= 84,

X3= 49.
C3=196, SUM=553

(I.e. x =2, x =3 and x3=4; the first output followed:
2
l

"le

OUTPUT:

Cl=273, C2= 90,

C3=l96, SUM=559.

c =42, c 2=77, c 3=90,L:c=209.,
l

* (The

Well, let us try an-

other type ••• (Interrupted by a new output: c = 42,
l
c V83 , C =90,L:C=2l5) ••• hmm, it's increased •••
3
2
O.K., let them be 45, 5., 4.

~utput

c l =60, c 2=3l,

(Le. x l =45, x 2=5, x3=4;

c3=698,~C=789.)

*
*
*
*

noise appears here.)
NEW POINT:
OUTPUT:

Xl= 4, X2= 2, X3= 0.
Cl= 36, C2= 98, C3=54l, SUM=675.

NEW POINT:
OUTPUT;

Xl= 5, X2= 48, X3= 3.
Cl=260, C2=102, C3=474, SUM=836.

Whoops, it's up ••• (interrupted by a new output:

c =60, c =29, c3=698,~C=787). Now, let them be
2
l
4, 48, 48. (Ou'tput: c =273, c 2=90, c3=754,~C=1117)*
1

')'(

Oh, that.' s bad.

0, 50, 0.

c3=460,~C=783.)

C2=104,

C3=474, SUM=838.

NEW POINT:

Xl= 24,

X2= 27,

X3= 22.

OUTPUT:

Cl=19l,

C2= 64,

C3=589, SUM=844.

rupted by output: c =27l, c 2=50,
l
Well, x =45, x =48 , x =4.
3
2
l
c=1184.)

Cl=260,

(Output: c =271, c2=52~
l

Still no good, let's mix them further ••• (inter-

c =879,
3

OUTPUT:

c3=460,~c=78l) •••

(Output: c =278, c 2=27,
l

This is the worst so far.

Now how about the center?

Say, 20, 25 and 30.

(Output: c =173, c =83, C3=6l7,l:C=873.)
2
l

,',
'i(

(p names nine points with coordinates either at the two ends
'i(

of the ranges,

°x

5 and 45

50, or in the middle,

')'(

30.)

')'(

*
~"
'i(

,,(
E:

How did you pick these values? Can you explain*

*
*
S: Well, I thought I would first try all these corners ••• a sort of local search at the two ends of every*
x range, and also in its middle. This might give me *
it now7

THE MATHEMATICAL LANGUAGE OF REPRESENTATION IS ACCEPTED.
INITIAL ASPIRATION LEVEL:

'i(

some idea of how that sum behaves ••• It could, of
')'(

course, go up and down in between ••• but that may
'i(

not be important.

I hope these functions (the hypo'i(

thesis is formed!)

do not oscillate too much •••
'i(

i'r

Ii
I~

Now I have ••• how many ••• four, five reasonable

points.

The sums here are no worse than, say, 800.

Let us be careful now ••• I want to give you a good Xl •••

,'r

*
"k

It's better if I don't care about these continuous outputs*
now ••• Well, when Xl was around zero, the sum was

about 210; when it was 28, we had almost 600.

-I~

So, how

about ••• howabout, say, if xlequals 10 ••• that might

hit the minimum ••• This kind of

,'r

th~ng

can give us, shall

we say, 15 for x 2 ' and ••• well ••• I'm doing the same for

x ••• O.K., let jt be 20.
3

,'t

736, Lc=778.)

(Output: c =134, c 2=108, c 3=
l

It didn't do much good ••• O.K., let us

include this point as well ••••

,..

SELECTED
Xl= 3,
Xl= 4,
Xl= 4,
Xl= 46,

X2= 47,
X2= 2,
X2= 3,
X2= 4,

*
,~

,'r

*
*
,'r
,'r

*
,'t
"Ir

,'r

,,,
*

X3= 49,
X3= 0,
X3= 50,
X3= 1,

SUM=553;
SUM=675;
SUM=680;
SUM=779.

. ones,
(p selects four points out of the nine " non-no~sy
with the lowest total cost values.)
INTERPOLATION:
INTERPOLATION:
INTERPOLATION:

Xl=2l;
X2=22;
X3=25.

(=3+(46-3).553/(553+779)
(=2+(47-2).675/(675+553)
(=0+(,50-0).675/( 675+680)

NEW POINT:
OUTPUT:

Xl= 21,
Cl=130,

X2= 22,
C2= 97,

X3= 25.
C3=563, SUM=780.

OUTPUT:

Cl=130,

C2=103,

C3=563, SUMf::786.

COUNTER NOW:

SUBSET OF POINTS:

(p fails to hit upon a sum better than at least 779, by
interpolation.

It is counting the number of failures.)

SELECTED SUBSET OF POINTS:
Xl= 3,
Xl= 4,
Xl= 4,
Xl= 46,
Xl= 21,
Xl= 5,

X2=
X2=
X2=
X2=
X2=
X2=

47,
2,
~,

4,
22,
48,

X3= 49, SUM=553;
X3= 0, SUM=675;
X3= 50, SUM=680;
X3= 1, SUM=779;
X3= 25, SUM=780;
X3= 5, SUM=804.

(p selects six points out of the ten "non-noisy" ones
so far, with'the lowest total cost values.)

I just can't get a better point ••• This is silly.*

Why don't I move off?

Suppose, we just try 35, 40,

45 for a change.

COUNTER NOW·:

)'r

RANDOM BREAK-OUT

*
*

ASPIRATION LEVEL NOW:

NEW POINT:

Xl= 17, X2= 38, X3=

(After five failures the aspiration level drops and
the next point is specified at random.)

,'r

(After hitting a point withl:c=58l)

Here we

are, that's reasonable ••• I see it now ••• This thing
seems to move downwards, toward the smaller x 's.
2
Let us choose ••• well, xl was all right, let it stay at

*
,'r

EXTRAPOLATION:
EXTRAPOLATION:
INTERPOLATION:

*
*

NEW POINT:

Xl=

Xl=O;
X2=48;
X3=23.
0,

X2= 48,

X3=

23.

"I(

8, ••• x 2 has to be down ••• down ••• say, 15 and x3
-k

stays in the middle, 25.

*
-/(

What is wrong?

,'r

I just don't get that sum any better.

I some"I(

how feel, I have to do something about the c's ••• I
mean, about c l ' c 2 ' and c 3 separately.
should minimize c

only, at first.

Maybe I

Let c

,'r

and c

Yes, that's what I am going to do.

Minimize them,

one by one, and then get them together again.

Where

ENTER STAGE (B).
POINTS WITH BEST Cl:
Xl= 4, X2= 2, X3= 0,
Xl= 45, X2= 4, X3= 46,
Xl= 32, X2= 11, uX3= 17,
Xl= 21, X2= 22, X3= 25,

Cl= 36;
Cl= 59;
Cl=106;
Cl=130.

(The difference between the inital and current
aspiration levels is four, which fact makes P quit Stage

whatever they are; and then come back to them •••

ASPIRATION LEVEL NOW:

*
*
*

(A) and enter Stage (B). P selects four points with the
lowest c values so far.)
l

,'r

was c

the best?

......
......

.......

,'(

-J(
~

-J(

Well, c

depends on x 2 and x 3 ' and a little bit,

maybe a little bit, on xl.

POINTS WITH BEST Cl:

And it also varies with time ••• *
-/(

Yes, low values of x 2 are all right •••

Xl= 4,
Xl= 18,
Xl= 28,

X2=
X2=
X2=

2,
5,
4,

X3=
X3=
X3=

2,
6,

Cl= 36;
Cl= 52;
Cl= 57.

.,~

(A further search produces these three best points

with regard to c .)
l

.,~

*
,'(
.,'(

You see, this was easy, to minimize c 2 ••• I now

POINTS WITH BEST C2:
.,'(

know what it is like ••••
.,~

~'~

-/(

Xl=
XI=
Xl=
Xl=

35,
45,
41,
36,

11,
X2= 38,
X2= 2,
X2= 43,

~2=

X3=47,
X3= 3,
X3=13,
X3=30,

C2= 39;
C2= 47;
C2= 51;
C2=53.

(The four best points w~th.regard to c .)
2

-/(

S:
it.

That c

is tough, I can't see much reason behind

Somehow, when I don't move with the

XIS,

it's small

*
*
-;'c

er, and when I change them a little, it jumps up ••• It is
a funny variable ••••

*
-/(
.,~

POINTS WITH BEST C3:
Xl= 3,
Xl= 27,
Xl= 6,
Xl= 11,
Xl= 3,

X2= 47, X3= 49, C3=196;
X2= 13, X3= 5, C3=343;
X2= 7, X3= 21, C3=352;
X2= 15, X3= 34, C3=355;
X2= 4, X3= 18, C3=374.

If I take, say, .36 for xl' 5 for x 2 and ••• and a

small value, say, 2' for x ' I should get just about the
3

minimum ••• I don't think I can do any better ••• not when
this noise is on all the time.
c =408, Lc=477.
3

End.
,'(

Output: cl~ 48, c =2l,
2

THE BEST POINT SO FAR:
Xl= 28, X2=

2, X3= 47, SUM=463.

(p was cut off here.)

,'(

-Jr
i'(

-J(

;'(

-J(

-Ir

*
The following points may be worth mentioning with regard to the comparison between the
above two records:
The simulation of both the results of and the reasoning behind the subject's decision making
is fairly faithful, although the trial points are, of course, not identical.

The only serious

shortcoming of the model can be seen at Stage (B), when it does not notice the effect of large
step sizes.

Consequently, the prograII).' s best points wi th regard to c

just, so to speak, happen

to be the best.

The first point on the list (x =3, x =47, X3=49 , c =l96) is "too good" and its
2
l
3
weight causes the final selection mechanism to choose a 'global optimum' with a much too large
X •
3

(The x3=47 value of the first point on the best-c -list was also guilty in this decision.)
2
The quality of the computer search for minimum was also very similar to the human one.

The

machine obtained a minimum of 463 after 68 trials, as contrasted with the subject's minimum of
477 after 59 trials.

S
i-o&

-.J

180

Errata

p. 821,
p. 822,
p. 824,
p. 826,
p. 827,

p. 828,

p. 830,

the display· characteristics for the associated
X and Y coordinate fields.
Fig. 3, last line: Change (short, med., long) to
(solid, dashed)
lines 40 & 44, L: Change 1278 to 1778
line 51, L: Change one bit to three bits
Table 2: Change left column pair heading to
Data List Before Delete
line 19, L: Change 1278 to 1778
lines 33-36, R: Change first sentence of paragraph to A typical delete subroutine also
consists of two machine instructions. Delete
remainder of sentence (lines 34-36)
line 24, L: Remove s from blocks
line 34, L: Add s to instruction
line 42, L: Change if to of
line 32, L: Change to the same central computer

A.BURNS
Hybrid Computer for Lunar Excursion •••
pp. 899-900, Figs. 3 and 4: Figures reversed (captions are correct)

A. P. SAGE and R. W. BURT
Optimum Design and Error Analysis • . •
p. 903, line 10, L: Change t to T
eq. 3: Change x(T) to x(t)
p. 904, eq. 10: Change y(n-1)T to y(ii=iT); and
y'(n - l)T to y'(n - 1 T)
p. 909, eq. 20: Change cosdT to cosbT

(11

K=O

eqs. 24 and 25: Should be
~ (nT) == \l yf[Yd (nT),

P] ~y (un T)
+ R[Ya (nT)-Yd (nT)]

(24)

Ap (nT) == \l J[Yd (nT)~
f] ~y-(n + 1 T)
+ Ap (fiT] T) (25)
line 1, R: Change 6 to \l
eq. 29: Change
!(jT»

~i(jT),

!(jT)

p. 912, eq. 21: Change I

M. W. GOLDMAN
Design of a High Speed DDA
p. 930, line 26, L: pIes equation should be pIer equation
line 13, R: ~ should be
p. 931, line 17, L: very should be vary
line 21, R: 1.000 should be 1000
p. 932, line 7, L: d~ should be d (~)
line 14, L: convential should be conventional
line 16, L: sole should be scale
p. 933, line 15, R: ~ X IT should be ~ X IT
p. 934, line 10, L: Replace all symbols u in eq. (1)
with u; Quantity f3r should be under square
root sign
line 12, L: Period after range
line 1, R: Replace all symbols u in eq. (2)
with u; delete comma and ft following integral
line 2, R: Replace all symbols u in eq. (3)
with IT
line 3, R: Replace all symbols u in eq. (4)
with IT; minus sign precedes quantity on
right side
line 6, R: Replace Ui with IT;
p. 935, Replace all symbols u with IT
line 8, L: mixi- should be maxiline 18, R: Quantity f3r should be under square
root sign
Fig. 2: Integrator marked cos t/J in upper
right is integrator 10
p. 936, Replace all symbols u with IT
line 3, L: Quantity f3r should be under square
root sign
line 18, L: 1Z should be liZ
line 24, R: Quantity x should be under square
root sign
p. 937, line 2, L: Ditto
line 5, L: Ditto and x == ~ should be x

+ to

line 3, L: d 1 should be d~
x
Fig. 3: chapman's in caption should be Chapman's
p. 938, line 11, L: th should be the
p. 941, line 1, R: Quantity ~ should be ~y'
"Registers and adders" should be italicized and
spaced as a heading
line 27, L: C
no overflow should be C no
overflow
yields should be ~
yields
line 30, L:
line 43, L: (C -OUT-) should be (C ==;:;::..
-OUT-)
bottom, R: follOWIng tabulation should be

<~i(jT),

p. 913, eq. 32: Change n q-H to n q+1
eq. 34: Change nq+1(nT) to nq+l(jT)

Errata

181

tabulation following Table 1
p. 943, Table 7: Should be replaced by
I

Condition of
Integrator

Input to Integrator
ax
aY
+
-

YM = 0; Y s = +

YM = 0; Ys =-

YM=#=.O; Y s = +
YM =1= MAX

YT +

Y M=1= 0; Y s = YM =1= MAX
Ys = +
YM = MAX

R~Q

+ ~Ys
YT + 1 ~Q

R~Q

I~Y

-~Ys

Yc + 1

QT~R

QT~Y

Qc~Y

C~+ OUTPUT

C~-OUTPUT
Y T + R~Q

Yc + 1 ~ Q

Y T + 1 ~Q

QT~R

Qc~Y

QT~Y

I~OVERFLOW

Yc +

R~Q

QT~R

Yc +

-~Ys

~Ys

l~Y

I~Y

~ -OUTPUT

C~+OUTPUT

YT +

Yc +

R~Q

QT~R

R~Q

C9+ OUTPUT
C:?-OUTPUT
Y c + R~Q
YT + R~Q
QT~ R
QT~R
C~-OUTPUT
C~+ OUTPUT
p. 946, Fig. 11: Add caption: Front and back view of
typical logic cards
line 8, R: comparisn should be comparison
Ys =YM = MAX

+ 1 ~Q

Qc~Y

QT~R

Yc + 1 ~ Q

l~OVERFLOW

Qc~Y

Z register
line 17, R: Close ORing

p. 947, Ref. 20: lEE should be IEEE
T. J. GILLIAN andP. B. PERSONS
High Speed Ferrite 2~ D Memory

p. 949, Ref. 51: 161 should be 1961
Ref. 52: Add Ref. 53 as follows: 53. H. J.
Weber, "Inertial Guidance System Uses Digital Integrator", Space/ Aeronautics, vol.
30, pp. 134-38 (Nov. 1958)

p. 1017: Graph in Fig. 11 should be interchanged with
photo shown in Fig. 15

M. RASPANTI
Training for the No. 1 ESS

J. T. QUATSE
STROBES: Shared-Time Repair ••.

p. 967, line 1, R: Change last months to last six
months

p. 1066, line 31, R: Change reduce to require
line 33, R: Change reduce to require
p. 1068, diagram (Test Oscillator), top center portion:
+ 12 should read - 12
p. 1069, line 1, L: Par should be PAR
p. 1070, bottom, R: should read
STROBES IS NOW PREPARED FOR MM11 E
T
p.1071, line 9, L: Should read T TSK, MEM, 2, 1;
line 15, R: Should be underlined STROBES
IS NOW PREPARED FOR MM11, E
last paragraph, line 2: modules should be
module
last paragraph line 4: modules should be
module
p. 1072, line 3, L: Deleted; should read procedures
which are essential to high system per-

p. 968, line 8, R: Change repeating backtracking to
repeating and backtracking
p. 969, line 4, R: Analogous misspelled
p. 970, line 40, L: Change approximate to appropriate
p. 973, line 2, R: Change respective to respectively
p. 975, line
line
line
line

38,
43,
44,
45,

R:
R:
R:
R:

Change X X Y to X X y*
Change X 0 N to X.J N*
Change of to by
Change X 0 N to X 4 N*

p. 976, line 4, L: Change W to f W Transfer to
line 13, R: Change x, y, and Z to X, Y and

Source Exif Data:
File Type : PDF File Type Extension : pdf MIME Type : application/pdf PDF Version : 1.3 Linearized : No XMP Toolkit : Adobe XMP Core 4.2.1-c041 52.342996, 2008/05/07-21:37:19 Producer : Adobe Acrobat 9.0 Paper Capture Plug-in Modify Date : 2008:11:17 07:16:31-08:00 Create Date : 2008:11:17 07:16:31-08:00 Metadata Date : 2008:11:17 07:16:31-08:00 Format : application/pdf Document ID : uuid:68d8d7f4-6cb9-4230-9900-7dfe11be9d94 Instance ID : uuid:8674e5f6-d7e2-471b-93df-ad5fc8385634 Page Layout : SinglePage Page Mode : UseOutlines Page Count : 199
EXIF Metadata provided by EXIF.tools

1965 11_#27_Part_2 11 #27 Part 2

1965-11_#27_Part_2 1965-11_%2327_Part_2

Navigation menu

Versions of this User Manual:

Views

Navigation