- -ADVANCED RESEARCH PROJECTS A.GENCY

Dec 22nd

MEMO FOR. __ P_r_o_f_e_s_s_o_r_M_c_C_a_r_t_h_y _________ _

Dr Sutherland said you said you

never received the copy of this I sent to

you under memo dated Nov 23, 1964, from him.

Marie Quinn, Secretary to Ivan E. Sutherland

-.~\[ 1"',:''':',­:. .;:=;;;..

_ .. :-;:---_. ------

;J~ ,t·:~.s-:ts~~·::,·~.:: tr;r::IJ;'~ ~l·Js;·/l

~·.:~.~S;".t5:~-jj~2 ,,~;> f s-7I

i":

.. -."

Goals

A RESEARCH AND DEVEIDPMEN'l' PROGRAM IN APPLICATIONS OF D'riLLIGENT

AUTOMATA TO RECOl'mAISSANCE

. The long-range goal of this program will be to develop automatons capable ot ga.thering Pl'ocessing and transmi tt1ns intormation- in " hostUe

'. . environment,_ The time period inVolved 1s 1970-1980 ~

The first short-range goal of the program w1ll be to design and develop a mobile automaton to aQcompllsh non.trivial missions 1n a real environment. ~ernaJ. control will be exercised over the automaton tram ,. computer.. The automaton .rill have at least visual and tactile aensor cape;bUit7.

The second short-f8:1lf5e goal will be to design and develop fit mobUe au­to!uaton to accomplish non-trivial missions in a real env1romnent 1D a solt­contained n-.ode, e.g •• with Uttle or no ~ernal. control..

, .;

other goals int~rmed.ie.ry to the long-range Boal will be pre41cated as research proceeds towards achieving the t1rst sbOrt-range goal. 'nlis f1rR ': . goa.l is anticipated to require three yeua. The lJocon4. ahort-r'ange project can-.·

:; ;' ~ initiated after about two year'a work on' the first Saal and will itself r .... ·e.·· quire three years giving a total duration of tive years tor t.cQompl1abment

of the f~st. two stated goals •

. ' • Such a long-range goal attained by .stepp.ding throUgh a number ot inter-.. med1a.ry ones 16 believed essential. to kn1t together 6$ man..v ot the constituent subje~t areas ot "artificial. intelligence" as possible. It bas been 80 stated as to require the successful application ot many techniques with aU the attendant 'problems of interaction and feed-back. It 1$ difficult of realization but by . the same token it w1ll prov1d8 801utl~ to e¥1at1ns presaiDs m111~ prOblems.

: 1wfethod of Approach :'~' .

': 1. Efforts to achieve the first sbort-range Soal 8h~ be r1n1t1.tecl . ;:lmmed1atel.y under the following constraints I . .1 ' . •

'. :" a •. : The mobile 4ev1ce should. operate in • real env1romnent' with , : ... :,' at least visual and t~lle sensora.· , . " " . ' .

. ' : .. ,;. b.·, It should be operateci Ui1der oomputer control and mq ,be caDle-I':. '. • " .~. " • -connected. ~. I •

1 '>:' I • 1 ~

" ;

;';

.. ;:Oi J . ;1"

. ". '. :': Q.: The "automaton . system" ~ODSi8t1Jlg ot' compute"i. molaU. autc:anaton t •

;';-' ' .. ana. coxmec'b1va Uldts·muat·bave.:tbe capabU1 __ wt: :-:-. .i • •..•. . ~ r .;.:; , . .' '-' ~!

• ;. : • ~ ~. I • ..,:' '~' ',.' • :,1

::~: .. :,~ L ;. .

), ,-

. , ..... - .. ': .. : :::. . ' . :~ ."1.

4 ~;:.'"

~ '. • I

' .• ' ,'! : .. '-'

" ; " I

• _I:" . .

'.' ' --.~~. ~ ,,'

::'. \ 't·'

'> ;;r(, : 1. : _:

" .,. " ','

.:. ~

" .. ~.

· ,;" ~ ~' ..

. :,:

-.... .'

I • ~

,~ .'. '-\ ': ..

· ';~. ': '

· ~',"

'.; .::

... : ." " ! •

. - ~. . :. • '. ~ • .' "'. '! ' ~ _. ____ .a...'_=_ ._~.: __ . ....:::.:.:...::._. ~..:.:....:.t._::::..:...-, ,:;.....,-~' . -'--_~j,._._. ---:.. ...... - ___ ...,...;......;;...,n-.,........"--~._.,...---,,...,.............,__.~ ....... ··.f";~:.

. " i

! j

j

I '

I I 1 ~ ,

.;.i ":-

., , "

2 -(1) Analyze the .~t 11& wb1ch ~be automaton tU48

"nU' ~'

(2) Determine the tasks necessar;y to perform ita ass1pecl , m1as1on 111 1te enviromaent

.... 1. (3) Pe~orm the tasks 1t baa set, up as e.aeDtia1 1;0 ita mission,· and

. : (4) Kodi.fY cont:1.nuoualy its behavior OD the b8l18 of 01>- ,

serve4 changes in :Lta enrirOJJmeJlt . " ..

. : ;' u: : ' ' d. 1'he bardwa.re of the autcxinaton w1l1 be kept to a m'n1xmma with .. ~ the analySis, task determ1DatioD aa4 system mo41t1cat1on beiDs performed b7 .: ]', 'the comput1Dg system • . . ;,

. :' ,~ ~ ,

.

.. : .:;. . ' ;;

.. :. ~ i .:

."

e. ~ comput1Dg s:rstem itself may conta1n spec1a1 pul'p088 equip- ' 'Jn6nt tor pattern reeosn1t1on, etc •

2. A few iniss1cms shall be cle8Z'~ 4el.i1leated ~ the J)()D anc1 the con-; tractor prior to contract 1D1t1atlon and interms4iate soa.1.a 8ball be emuneratecl •

.; '2:',. A typical hliasion wu14 be to collect tzom an 38acq.-tment of objects , I .:,:, ;; '.:, ,;, . in a given room those hav1Jls prescribed pbyB1cal eha:cacter1at1c8 and to collGct. ' I ' .. :'.::',::" " them in a given area. Another typical. m1aa1on voul4 'bG 1;0 arranse a set of ~ , :.:;. " boxes 80 as to move to a sinn po1Dt e1tber ~ the WI8 of the 'bQxes 0'"

.'

. .'

;,.,~ ... _~:~ __ ~_a;_ll_~~1r a~~~~_n_ce_,_or_~~~"!M! __ "-'--.... __ '_' ____ . _____ ---',:-: ,._' ~,_i~,,,,-__ ~~~,-:-' :._,. __ , ~

j ,:,

; .

, .. , ,. '

;; ,',

L.~' :.

~j.::,' , ~, ..... ' .. :':: . .. ~,.;. ': t·;' h· . r ' ~; ..

1'<·", ,,_,'. :i -F ;,': :r".: '1-, :,/. ~.

'J .1 ••

~ .

~t2 I 7i1so dt!?JJ- -rf

ADVANCED RESEARCH PROJECTS A.GENCY

Nov 23, 1964

MEMO FOR Prof John McCarthy - Stanford

This proj ect is getting started and

we are getting a proposal for the work

outlined here. Expect this will fit well with

your new laboratory proposal.

Ivan E. Sutherland

'.-:' .... f ," ~ ')" .~. ,- - ~. '" . . '.

r. ' -' - . -. :~ s :.::.; :.:.::"~ ':"',

...... :-

:> -••

Goals

A RESEARCH AND DEVEU>PMENT PROGRAM IN APPLICATIONS OF INTELLIGENT

AUTOMATA TO RECOrniAISSANCE

, The long .. ra!J.Be goal of this program wUl be to' develop autOJnatons capable of gathering processing and transmitting information in a hostile environment,_ The time period involved is 1970-1980.

The first short-range goal ot the program will be to c).esign and develop a mobile automaton to accomplish non-trivial. missions in &, real enviromnent. External. control will be exercised aver the automaton trom a computer. The' automaton Will have at leaSt visual and tactile sensor ca.:pabUi ty.

The second short-range goal' will be to design and develop a mobUe au­tomaton to accompJ.1sh non-trivial missions in a real environment in a aelf­contained DiOde" e. g., with II ttle or no external control.

other goals intermediary to the long-range goal will be predicated as research proceeds towards acbieving the first short-range goal. This first gO.'i.l is anticipated to require three years. The second short-range project can be initiated. after about two year's work on the first goal and will itself re­quire three years giving a total duration of five years tor accomplishlnent of the first two stated goals.

Such a. long-range goal. attained by stepping throUgh & number of inter­mediary ones is believed essential to knit togethe~ ae many ot the constituent subject m,'"eas of ua.rtific1al intelligence rt as possible. It has been so stated as to require the successl'ul application of many techniques with all the attendant problems of interaction and feed-back. It is difticult of ~aJ.1za.tion but by the same token it will provide solutions to existing pressing military problema.

Method of Approach

1. Efforts to achieve the first short-range goal should be initiated irmnediately under the following constraints:

8.. The mobile device shouJ.d operate in a l:'eal environment' with at least visual and tactUe Gensors.

b. It should be operated under computer control and may be cable-connected. '

Q. The "automaton system" consisting of computer, mobile automaton and connective units must have the capability to: -

, .. ~

. - -. .. .. - ----:---'------.,-~'---- -----=- ----- --------------- ---.4 -

,/

) i I I j

I I I I I I

I I I I

I I !

po" ~ .-' ~,

2 -(1) Anal.yze the environment 1n which the automaton rinds

itself .

(2) Determine the tasks necessary to perform its assigned mission 1u ita environment

(3) Perform the tasks it has set up as essential. to its mission; and , \

(4) ModifY continuously its behavior on the basis of ob-served changes in its e%XViromnent

d. The hardwa.N of the automaton will be kept to a minimum with the analysis, task determination and system modification be1.xlg performed by the computing system.

e. ~e comput1Dg S1stem itself may contain special purpose equip­ment for pattern ree~t1on, etc.

2. A few missions shaJ.l be clearly delineated by the DOD and the con­traetor prior to contract initiat10n and intermediate goa.1.a shall be enumerated.

A typical mission would be to collect trom an aBsortm2nt of objects in e. given roam those bAv1ns prescr1bed physical characteristics and to collect them in a given area. Another tY,plcal miasion would be to arrange a set ot boxes 80 as to move to a. given point either tbrough the use of the boxes or tbrougll their avoida.uce, or both. . .--- • __ .....J ___ .._ ___ • _____ •• __ ~ __ _ -"~)-- ..

•

"

ReqUest for Supplementary Support

of

Research in a Time-Shared Oomputing System

(present grant GP-3207)

Principal Investigator

John McCarthy

Protessqr of Computer Science

Computation Oen~er

~tanford University

Sept~mber 10, 1965

-.~

Request for Supplementary Support

of

Research in a Time-Shared Computing System

This is an application for a supplementary grant of $60,000. to our

NSF grant GP 3207 to enable us to improve the Stanford time-sharing system

whose development is now being supported by NSF. Stanford University expects

that the indirect cost rate will have to be negotiated; and the 20% rate is

used here to establish the upper limit.

The amendment will make possible the following improvements.

1. To expand the number of teletypes from 8 to 16 allowing us to locate

consoles in a number of departments.

2. To expand core memory from 20,000 to 32,000 words. This will allow

new user services such as a version of JOSS by providing additional system

subroutines such as floating point. It will allow prompt service to c~rtain

non-swapped time shared users.

3. To provide a printer capable of printing the 114 character font on

our own display units.

The amount requested is half the cost of these improvements. The other

half will be borne by the Institute for Mathematical Studies in the Social

Sciences that shares the computer with us.

Present Status of the Stanford Time-Sharing Project.

We are now one year into the two year time-sharing project. The following

has been accomplished so far:

1. The ODIN preliminary time-sharing system was put ~n operation in

August 1964 and improved during the fall and winter. It serves five users at

1

once, has an external connection to the teletype networks, uses the Philco

displays as in-out devices, and uses mainly pre-time sharing PDP-l utility

programs.

2. A system that allows the use of the IBM 1301 disk for storage was

completed in January 1965.

3. An algebraic compiler called Gogol was completed in February 1965.

4. A text editor (TREDIT) using the displays was completed in March 1965.

5. A system that allows 7090 jobs to be initiated from the PDP-l was

completed in November 1964. This includes facilities for putting PDP-l files on

the 7090 output tape for 1401 listing. 7090 program in BALGOL and LISP may

be run.

Programming languages available on the PDP-l include MACRO, LISP, and GOGOL.

The ODIN system is quite bug-free at present and is in use 18 hours per

day. It can be used via the TWX network, and several outside users have done

so.

The late delivery and long debugging period of the Philco displays took

a lot of time from other tasks. However, the final result is quite impressive;

for example the hardware performance is better than similar IBM hardware promised

a year from now and the cost is less. The presence of the displays has induced

a number of interesting applications and more are expected.

Plans for the Next Year.

The new time-sharing system is about 99% coded and 95% debugged. It has

the following advantages over the old system:

1. 24 users at once rather than 5.

2. The 12 display units can be used as consoles.

3. A switchboard input-output system th~t allows any program to talk to

any device.

2

4. New user services such as an on-line disk calculator and ALGOL

interpreter, ~GOL, and a programming teacher.

5 • An ilnproved command language.

6. A flexible set of subroutines and commands for controlling displays.

1. A new 1090 monitor that will allow time sharing via the PDP-I.

8. A new assembler and a new text editor.

The basic system will be completed in October and'the remaining months of

the project will be denoted to the development of user services.

The Need for Additional Hardware.

Both the Computation Center and the teaching machine project would like

to expand the system in three respects:

1. More core memory_ We wish to add three more blocks of 4096 words to

bring the memory up to 32186. This will enable

Personnel.

a) floating point routines to be added to the system aiding

on line ALGOL.

b) conversion routines to be added.

c) a collection of new display services to be added.

d) a time sharer that is not swapped from core for high-speed

interaction experiments.

Professor John McCarthy is the Principal Investigator of the Stanford

Time-Sharing Project.

He has the following people working on the project:

Stephen Russell, John Allen, Margaret Novotny, Gary Feldman, Programme~'Analysts;

David Poole, John Sauter, Student Research Assistants;

Engineering Techniciano

3

Jordan Benedict,

Working closely with the above group, which is developing the basic

system, are Dow Brian, Programmer/Analyst; Paul Stygar and Brian Tolliver,

Student Research AssistantsD These three are employed by Professor Suppes'

teaching machine project to adapt the time-sharing system to this latter major

project.

No salaries are included in the budget of this proposed supplement

as there are adequate funds in the basic grant for the personnel commitment.

4

BUDGEI'

The following are the estimated costs of the items required.

1. 12,000 additional words of core

2. Digital Equipment Corp. PDP-8 line scanner with 16 teletype interfaces

3. 16 Mode 33 KSR teletypes

4. Analex line printer with interface 200 lines/minute 0 128 characters

5. Interfacing electronics

California Sales Tax

6. 1 year maintenance at 5%

Total direct cost

7. Indirect costs (20% of direct)

Of this the teaching machine project will pay

half leaving

which we are requesting.

5

$

$

$

$

$

30,000.

25,000.

14,400.

25,000.

5,000.

99,400.

3z276. 103,376.

4,970.

108,346.

2lz669.

130,015.

60,000.

Proposal submitted by:

John McCarthy Professor of Computer Science

Approved:

Edward Ao Feigenbaum

Director, Stanford Computation Center

Hubert Heffner

Vice Provost for R~search

6

,

Request for Supplementary ~upport

of

Research in a Ti~e~Shared Cqmputing System

(present grant GP-3207)

Principal Investigator

John McOarthy

Professqr of Comput~r Science

Computation C~n~er

~tanfor.d University

September 10, 1965

Request for Supplementary Support

of

Research in a Time-Shared Computing System

This is an application for a supplementary grant of $60,000. to our

NSF grant GP 3207 to enable us to improve the Stanford time-sharing system

whose development is now being supported by NSF. Stanford University expects

that the indirect cost rate will have to be negotiated; and the 2~ rate is

used here to establish the upper limit.

The amendment will make possible the following improvements.

1. To expand the number of teletypes from 8 to 16 allowing us to locate

consoles in a number of departments.

2. To expand core memory from 20,000 to 32,000 words. This will allow

new user services such as a version of JOSS by providing additional system

subroutines such as floating point. It will allow prompt service to certain

non-swapped time shared ueers.

3. To provide a printer capable of printing the 114 character font on

our own display units.

The amount requested is half the cost of these improvements. The other

half Will be borne by the Institute for Mathematical Studies in the Social

Sciences that shares the computer with us.

Present Status of the Stanford Time-Sharin~ Project. i

We are now one year into the two year time-sharing project. The following

has been accomplished so far:

1. The ODIN preliminary time-sharing system was put in operation in

August 1964 and improv~d during the fall and winter. It serves five users at

1

once, has an external connection to the teletype networks, uses the Philco

displays as in-out devices, and uses mainly pre-time sharing PDP-l utility

programs.

2. A system that allows the use of the IBM 1301 disk for storage was

completed in January 1965.

3. An algebraic compiler called Gogol was completed in February 1965.

4. A text editor (TREDIT) using the displays was completed in March 1965.

5. A system that allows 7090 jobs to be initiated from the PDP-l was

completed in November 1964. This includes facilities for putting PDP-l files on

the 7090 output tape for 1401 listing. 7090 program in BALGOL and LISP may

be run.

Programming languages available on the PDP-l include MACRO, LISP, and GOOOL.

The ODIN system is quite bug-free at present and is in use 18 hours per

day. It can be used via the TWX network, and several outside users have done

so.

The late delivery and long debugging period of the Philco displays took

a lot of time from other tasks. However, the final result is quite impressive;

for example the hardware performance is better than similar IBM hardware promised

a year from now and the cost is less. The presence of the displays has induced

a number of interesting applications and more are expected.

Plans for the Next Year.

The new time-sharing system is about 9~ coded and 95'fo debugged_ It has

the following advantages over the old system:

1. 24 users at once rather than 5.

2. The 12 display units can be used as consoles.

3. A switchboard input-output system that allows any program to talk to

any device.

2

4. New user services such as an on-line disk calculator and ALGOL

interpreter, ~GOL, and a programming teacher.

5 • An llnproved command language 0

6. A flexible set of subroutines and commands for controlling displays.

7. A new 7090 monitor that will allow time sharing via the PDP-I.

8. A new assembler and a new text editor.

The basic system will be completed in October and'the remaining months of

the project will be denoted to the development of user services.

The Need for Additional Hardware.

Both the Computation Center and the teaching machine project would like

to expand the system in three respects:

1. More core memory. We wish to add three more blocks of 4096 words to

bring the memory up to 32786. Thi s will enable

Personnel.

a) floating point routines to be added to the system aiding

on line ALGOL.

b) conversion routines to be added.

c) a collection of new display services to be added.

d) a time sharer that is not swapped from core for high-speed

interaction experiments.

Professor John McCarthy is the Principal Investigator of the Stanford

Time-Sharing Project.

He has the following people working on the project:

Stephen Russell, John Allen, Margaret Novotny, Gary Feldman, Programme~'Analysts;

David Poole, John Sauter, Student Research Assistants;

Engineering Techniciano

Jordan Benedict,

,

Working closely with the above group, which is developing the basic

system, are Dow Brian, Programmer/Analyst; Paul Stygar and Brian Tolliver,

Student Research Assistantso These three are employed by Professor Suppes'

teaching machine project to adapt the time-sharing system to this latter major

project.

No salaries are included in the budget of this proposed supplement

as there are adequate funds in the basic grant for the personnel commitment.

4

BUDGEI'

The following are the estimated costs of the items required.

1. 12,000 additional words of core

2. Digital Equipment Corp. PDP-8 line scanner with 16 teletype interfaces

3. 16 Mode 33 KSR teletypes

4. Analex line printer with interface 200 lines/minute. 128 characters

5. Interfacing electronics

California Sales Tax

6. 1 year maintenance at 5%

Total direct cost

7. Indirect costs (20% of direct)

Of this the teaching machine project will pay

half leaving

which we are requesting.

5

$

$

$

$

$

30,000.

25,000.

14,400.

25,000.

5,000.

99,400.

3,276 •

103,376.

4,970.

108,346.

21,669.

130,015.

60,000.

••

Proposal submitted by:

John McCarthy Professor of Computer Science

Approved:

Edward Ao Feigenbaum

Director, Stanford Computation Center

Hubert Heffner

Vice Provost for Research

6

i·

. .

'a •• '~

ADVANCED COMPUTER FOR MEDICAL RESEARCH

Stanford University

This proposal is submitted on behalf of Stanford University School of Medicine with the aim of furnishing a computation facility that can match the other dimensions of research capabilities and facilities of the s~hool.

The sequence of our presentation is arbitrary, and reviewers are urged to undertake a detailed scan after a quick pass to map the salient blocks. ~ refers to the Stanford Computation Center. ACME is an acro~ for Advanced Computer for MEdical Research.

Outline

(1) Source of the proposal. Medical School Computer Policy Committee.

(2) Medical School - University relationships. ACME will be an advanced facility to complement, not compete with SCC.

Health research at Stanford - Medical School.

(3) Present services of SCC. IBM 7090 and Burroughs 5500. Efficient job shop but language incompatibilities; limited file service, A-D, pro­gramming and engineering support.

(4)

(5)

(6)

System efforts on small computers now running: ·4 LINC's, 1 PDP-8 • Further prospects in data acquisition and process control.

Automated Biological Laboratory - NASA supported program for compre­hensive system study of automation of biochemical experiments.

The ACME proposal: dynamic complementarity. Operating policy, starting Spring 1966. Multitasking under Operating System/360. Adjustment of priority schedules for multiple users, telecom, file service; long jobs; high data rate interactions.

Hardware: IBM/360-50 and an 1800 satellite, process-control computer. A-D and D-A I/O gear. Lines to 20+ peripheral typewriters and data links. Central display consoles and plotter.

(7) scc-67 plans - starting Spring 1967. Switchover for time-shared use, releasing ACME for special applications. Policy convergence.

(8) ACME management and staff. Deputy Director from SCC for technical management under policy committee direction. Technical support staff accounts for half of budget.

(9) Responsibilities of policy committee; policy guidelines and preliminar,y time table.

(10) Some applications, not an exhaustive list, but sUbmissions mainly from policy commdttee members. These go from neurophysiological data correlation to vital statistics to mechanized inference for analytical biochemistry. .

a. Biochemistry (Stryer)

b. Pharmacology (Killam)

c. Genetics - Demography & Instrumentation (Lederberg - Bodmer)

d. Computer Science - Mechanized Induction (Feigenbaum)

e. Neurology (Morrell)

f. Anesthesia (Bellville)

g. Scientific Information Retrieval

(11) Housing the computer. Initially in present Medical School building. Potential for adjacent quarters juxtaposed to SCC.

(12) Interim support during staff buildup. (Engineers from Instrumentation Research Laboratory. System programming from SCC. Macy Foundation planning grant.)

(13) Curriculum vitae of committee members.

Bibliography of publications from Stanford Medical School illustrating computer applications.

(14) Budget.

(1) Source of the proposal.

The initiative for and endorsement of this proposal come from the computer policy committee, and ~he draft text from its chairman. Members of the Computer Science Department and Stanford Computation Center (SCC) have then been consulted on technical issues and for agreement on the management responsibilities. A larger interdepartmental users' group and the Executive Committee of the Medical School have likewise been informed and consulted, and Dean Glaser has played an active role in developing these plans. Thus every effort has been made to frame a proposal that would reflect the requirements and interests of the entire medical school, as well as the long range interests of Stanford University.' The policy commdttee will continue to act in this repre­sentative role as there are now too many competent users for them all to sit at once on a briskly functioning committee. Regular meetings of a users group will however be continued. .

The membership of the policy committee* is:

Dean Robert J. Glaser (ex-officio)

Prof. Lincoln Moses (Statistics and Preventive Medicine)

Prof. John W. Bellville (Anesthesia)

Prof. Lubert Stryer (Biochemistry)

Prof. Keith Killam (Pharmacology)

Prof. Frank Morrell (Neurology)

Prof. Edward Feigenbaum (Director, Stanford Computation Center)

Prof. Joshua Lederberg (Genetics), Chairman

*appointed jointly by the Dean of the Medical School and the Provost.

(2) Medical School - University relationships.

Stanford made a calculated decision to move the medical school from San Francisco to the Palo Alto campus, and did so in 1959. This was a con­scientious self-dedication to far-reaching communion of medical research and education with university life, based on the principle of mutual interdependence of medicine with the physical sciences and with human affairs. In future out­look, computers will be central to communication within the University, and this motive alone reinforces a policy of the greatest feasible convergence with the central facility for the University, serviced by SCC. In addition, there are important economies in avoiding redundant programming systems and languages, maintaining common libraries and files as well as in using the most sophisticated hardware. However, medical research does have special requirements and chal­lenges and there is bound to be a significant gap, at least in time, between the level of service that the SCC can offer the University as a whole and the

3

technical possibilities of the art. The gap arises 'in part from the specialized requirements of medical research (in some measure the possibility of leapfrogging over a good deal of analogue hardware that should have been developed, but has not been); in part from the special impetus and financial support enjO,yed by medical research, whi ch after all conveys the purpose of a distinctive Nm pro­gram; in part from the greater flexibility with which a service to a more coherent group can be administered compared to one announced campus-wide.

The purpose of this proposal is not to compete with the sec but simply to fill that gap, i.e., to complement sec services. To a very large extent, ACME will be a computer for research and development in computer techniques in health sciences; insofar as these techniques become reliably available on sce service, they will be withdrawn from the ACME schedule. In view of the rapid anticipated growth of demand, this is quite essential if ACME is to remain available for further systems research and special applications. This policy is also consonant with the view that production-type services should eventually be charged as current expenses to individual research grants, a limitation which would be stifling for development work on the computer systems themselves. An outline of SCC services is appended (Sections 3 and 7).

As will be noted, the urgent gaps ACME will face are expected to evolve with time: those now most evident are a general time-sharing and file access system; priority batch processing, high data rate (710kc) closed loops, and symbiotic interactions of the computer with live experiments.

(3) Present services of SCC.

The Center, located only a few hundred yards from the medical school, operates a job shop for the campus, serviced by what is now full time service on two computers: an IBM 7090/1401 with SUBALGOL, LISP and FORTRAN, and a Burroughs 5500 with ALGOL. For such a shop, the Center has an excellent rep~­tation: for example, actual performance of three express runs daily on the IBM 7090, with deck-in to program-out time less than 90 minutes for 1 - 2 minute runs being the rule. Turnaround on the B-5500 is often much faster. Part of this performance is attributable to the local monitor and SUBALGOL compiler which were written at the Center.

On the other hand, some justifiable complaints may come from the unwil­lingness or inability of SCC to offer complete problem-solving services. From a user's standpoint the utter incompatibility of languages on the two computers appears like a calculated source of frustration. Until recently, even data tapes were mutually unreadable with no recourse.

Programmdng assistance for file manipulation is difficult to find, and many users have not had enough incentive (or resources) to recruit full time staff for their individual needs.

Attempts to operate direct wire communication from a LINC computer to the IBM 7090 (for interchange of data and programs via disk files see Section 4) will have borne fruit only after two years of intermittent effort. Not that this is an easy task for the 7090, but the main problem has been the preoccupation of sec systems staff with other, perhaps generally more urgent concerns.

SCC has been unable to offer A-D conversion service, not for lack of the suggestion. It has, however, pioneered in the routine availability of plotter outputs from a job shop, supported by excellent programming packages.

The grievances are recited not in malice, but to point out the kind of initiative that must be mounted at the medical school for the full realization of the enormous po~ential value of the basic services from SCC.

In any case, the 7090 is rapidly obsolescent in competition with the larger, faster, more versatile, and above all, more accessible machines of the next generation. However, SCC now plans to retain the 7090 at least through Spring 1967. Indeed, its displacement by the next central system, a time-shared IBM/360-67, will not be a completely unmixed blessing (except for FORTRAN stalwarts) since SUBALGOL will probably be discontinued and there will doubtless be some delay in the reconstruction of program packages, especially for statistical analysis.

SUbstantial use of the 7090 by the medical school must therefore be expected to continue, ACME notwithstanding, as long as SCC maintains it •. Indeed for an interval we may envisage some users conditioning their data on ACME to a certain point, then moving to the 7090 to exploit existing programs.

The administration of SCC has recently been strengthened by the appoint­ment of Professor Feigenbaum as Director, in preparation for the broader services established for the next generation computer. This promises much more effective communication on the problems of mutual concern of SCC and the medical school.

(4) System efforts on small computers now running.

Stanford was perhaps unique in obtaining two machines under the LINC evaluation program. One of these (a) was funded independently by Professors Chow, Killam, Morrell and Pribram for neurophysiology, and the other (b) was allocated to Professor Lederberg for general instrumentation. Subsequently, a third and fourth LINC (c) have been purchased for additional neurophysiology work, and recently a PDP-8 (d) by Professor Pribram to monitor behavioral studies. In spite of the software limitations, the machines are signed up fully for the day shifts and have more than a little tie-up for odd hours.

Applications have consisted of a fairly versatile series of programs making maximum use of the relatively limited capacity of the LINC and PDP-8 computers. This has been invaluable for preprocessing of data and rapid assess­ment of various computational techniques as applied to a wide range of biological problems. Actual applications have been listed in Section 10 for the most active groups, and the list is by no means exhaustive.

The Instrumentation Research Laboratory had special needs for versatility in new programming and its programming group wrote an operating system to generate a BALGOL-like compiler (BLINC), a new assembler with deeper diagnostics, a linking loader for library utility routines (kept, with the monitor, on microtape) which include several I/O'and displays - teletype, plotter (out) and curve-reader (in), scope display, and an IBM-compatible tape drive. The compiler runs on the IBM 7090 at SCC, being a revision of the SUBALGOL compiler to generate LINC code, which is

written on tape for communication to the LINC. With some bootstrapping, the rest of the system was written in BLINC; after compilation on the 7090, it is maintained on the LINC. This has motivated a long-standing, finally dubiously successful effort to communicate from the LINC to the 7090 by phone wire.

The utility routines have also incorporated multiple precision floating point arithmetic which has been made available to a statistical desk-calculator program. This keeps many students and department staff waiting in line for jobs like n x m X2 •

Other typical applications are curve-reading to digitize strip-chart recordings from a gas chromatograph and calculate areas of marked segments. This has so far been more economical and manageable than analogue tape recording.

The experience has been educational, but tantalizing, the LINC being barely capable for pointing up the possibilities of computer interaction with a variety of instruments. LINC programming software is still primitive, and if it were not, we would soon lack for enough copies of the machine unless it could be time-shared.

(5) Automated Biological Laboratory.

This program, under Professor Lederberg's direction and NASA support, is intended to be a convergence of medical research methodology and mission requirements for biological exploration of the planets. Given the high per­formance capabilities of the Saturn V launchers (e.g., 50,000 Ibs. soft-landed on Mars) of the mid-70's, we seek to emulate in an automatic laboratory a broad and flexible opportunity comparable to work in a terrestrial facility. This in effect demands the capability to program and reprogram an experiment in biochem- . istry or microbiology. While remote operation is not often so vital for medical research, many other aspects of laboratory automation would be extremely useful. Indeed, they may be quite indispensable for programs like health survey work for biochemical idiosyncrasies, or the chemical synthesis of functionally important polypeptides and polynucleotides. We believe further that many less exotic uses of intruments could be augmented by routine access to computation. In fact, many instruments that are now prohibitively expensive, or do not reach theoretical potentials of speed and sensitivity, would become available by the use of programmed logic on a time-shared general purpose computer in place of special purpose analogue hardware. The design of an integrating (and transfer-fUnction-correcting) photo­densitometer or microspectrophotometer, or of an interferometric IR-spectrometer shows this clearly.

An Instrumentation Research Laboratory, with a professional and senior engineering staff of some eight people currently, has been established, mainly with NASA support at present, with these stated objectives:

(a) General system study of an ABL for a Mars landing in 1975 or so.

(b) Some specific experiments for detection and characterization of exotic organisms.

(c) Find maximum advantage of this technology for analogous problems in biomedical research.

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

Each of these objectives is highly relevant to ACME.

(a) We have proposed to use the existing functional operation of the Genetics Department and some cooperating laboratories as a prototype. If this proposal is accepted, a number of experimental procedures will be engineered for automated operation under time-shared computer control. Most important will be the system-design study of the information and control traffic of such a set of diverse experiments with unpredictable demands for each next step. IBM Federal Systems Division is also collaborating on this study.

(b) The most exciting of these is the sensitive detection of optically active molecules by mass spectrometr.y and gas chromatography. A related project is the scanning of a specimen with a microbeam to elicit a mass spectral finger­print of each picture-point of a specimen, say a set of chromosomes. Programs for the automated analysis of mass spectral data are being pursued for the reduc­tion of this encyclopedic information.

(c) NASA policy encourages parallel emphasis on terrestrial applications • . Its support of the LINC eValuation program is a historic example. We can reason­ably expect continued backup from NASA for our instrumentation laboratory as an important complement to NIH-supported work.

ACME should, however, be distinguished as a general medical school resource, of which the Instrumentation Laboratory will be only one among many users.

7

(6) The ACME proposal.

ACME is not a single machine but a continuous adaptation of the medical· school to evolving requirements and technology in computat~on, on the one hand, and the tested production services that can be purchased from the campus system on the other. The opportunity to lease the hardware so that specifications can be changed on short notice is an important element of flexibility. The avail­ability of an expanding series of program-compatible machines is another. Finally, it rests on the basic policy that users should buy services from the central system wherever this is technically justifiable. This has two crucial advantages: the econo~ of scale with the strengthening of intra-university communication by centralizing well-established modes of services,and also releasing the staft, budget and hardware ot ACME as tar as possible for experimentation in newer and untried modes of operation.

The IBM/360-50 has been selected for the initial realization of ACME (1) as a machine technically appropriate to the immediate tasks in mind, and (2) for its system compatibility with the 360-67 alrea~ selected for the eventual replacement of the 7090 by the Stanford Computation Center. The 360-50 will be installed in ACME May 1966 and will run on three shifts under Operating System/36o, subject to review by the policy committee. These will be dedicated respectively:

(A) A prompt access time-sharing mode - perhaps over most of the working day.

(B) A scheduled, full-use, on line mode - to service development work on high data rate and on line control applicatons, and for similar systems development.

(c) Job-shop, especiallY longer runs for which overnight turnaround is acceptable, and which cannot be serviced with comparable etfective-ness by SCC. .

These functions in fact are represented by alterations in the Supervisor program ot Operating System/360, being mainly the reallocation of priorities for service under it.

The outline of the computing environment that will be available to ACME users might occupy a substantial part of this text. However, facilities of Operating System/360 have been outlined in great detail in a series of IBM publications which are readily obtainable (IBM Operating System/360: Introduction; Concepts and Facilities; and further reterences, especially, Job Control Language; Telecommunications; Data Management; Fortran IV, PL-I,and Assembler Language; also IBM System/360 Summary and Principles of Operation}. Since we intend to adhere closely to Operating System/360 for ACME, we can save unnecessary padding by reference to these publications. Job control specifications may well be incompatible as between ACME and the SCC, but they should represent a small burden, while ~ser programs and data set references should remain tully compatible.

To implement (A) a network of twenty typewriters will be installed (in general, one for each department) • Additional lines will be available for further users, and some 35-50 are expected, presumably ~ 32 active at one time, at this stage. An equal number of data lines will accommodate information from the m~ instruments that furnish less than 4 cps output. 'This network will teed a data bank, each account of which is recallable by name by the originator.

-~--~.+,-.~~-. .. .... --_._ ......... ~"--- ----.. .

Mode (A) will support most program-writing and debugging, information­retrieval, and data-management operations. It will also cover a substantial portion of production runs as the background jobs, subject to interruption under the time-sharing system. An important aspect of this as a system experi­ment is the level of service that is established for the longer jobs, on which a legitimate and controversial uncertainty now persists (i.e., whether their com­pletion will be intolerably deferred by cyclic service to the mix of short jobs if these always have higher priority).

(B) Some examples of high data rate work which is now rather frustrated are multiple channel electroencephalography, mass spectrometry, and video inter­pretation, e.g. for fluoroscopy and scintillography. Eventually it is hoped to develop more economical ways to deal with these situations, either with small peripheral computers or with high speed channels, but itwould be very painful to work these out on a computer so dedicated to uninterruptable general service that this work cannot proceed at a high priority.

(c) The job shop run of the mill is probably typical of any school with active programs of physiological research, as well as a number of demographic ' and epidemiological studies. However, some clinical research applications of electroencephalographic spectral analysis m~ require long runs with reasonably prompt turnaround, which it is not immediately obvious can be serviced adequately without special priority.

Besides the peripheral lines and the main frame, a comprehensive data interface will be installed at the computer. The detailed design of this is still under study (See Section~)particularly the pros and cons of an 1800 satellite computer vs. an 1827-4600 data control and high speed multiplex or channel. This would service a number of digital and analog input and output lines and relay registers. A tentative configuration is outlined for budget purposes.

ACME INITIAL CONFIGURATION: IBM/360-50 •• 1800

/360-50F processor, 65K bytes main core; direct control; multiplexer channel; 3 selector channels

large core storage 1 megabyte data cell drive 400 megabyte 2 (62311) disks transmission control; 20 (62741) terminals 4 tape drives (2 with 7 track operation) card read punch; paper tape

1801 processor 8K words main core adapter for S/360 1 62310 disk

The 1801 is intended mainly as a programmable high-speed multiplexer channel, analogous to the 1827-4600, also being considered.

Data channels, A-D, and other front end hardware details are still under stu~. This includes RPQ studies for high speed A-D units; channel logic for averaging, and microprogram hardware for direct access to external lines as addressable pseudo~emory.

(7) scc-67 Plans.

Starting Spring 1967, i.e., in ACME's second year, SCC will install an IBM/360-67 as a central campus computer supported by users' fees together with other 'funds. It will be operated under the now generally familiar IBM/36o-67 time sharing system, like a number of other universities~ As the SCC reaches an acceptable standard of service of various types, that mode will be discontinued on ACME, as determined by the policy committee. From the outset, ACME consoles will have the option of being switched to the /360-67_

The release of ACME from the otherwise high-priority demands of general time sharing will of course leave it available on a much more flexible basis for special applications. This will, of course, be the occasion for reconsid­eration of the most appropriate hardware. At the ver,y least, important economies could be won by sharing files and routine I/O equipment.

The depth and reliability with which this /360-67 system can service data-oriented users is a subject of lively controversy. We expect to have an answer to this question without the penalty of the frustration of experimental progress that a miscalculation would impose.

The policy guidelines and a preliminar,y time-table are presented in Section 9.

(8) ACME Management and Staff.

The facility would be under the direction of a computer policy committee' (designated in Section 1). This consists of medical faculty representatives with the participation of the Director of the Stanford Computation Center (Professor Edward Feigenbaum). The technical management of the computer, par­ticularly with reference to system programming, will be delegated to an associate director of the Center. This device is intended to minimize the duplication of system efforts on campus, and is similar to the arrangements already in effect for the very large SLAC (linear accelerator) computation center. ACME will be staffed by some two or three system programmers, three or four general applica­tions programmers, and two "hardware engineers", in addition to operators, dispatch assistants, etc. We have a strong university department o~ computer sciences, and especially close relationships with Professors McCarthy and Feigenbaum, and our own established interest and competence in a variety of computational techniques. Hence the appointment of a new specifically computer­oriented faculty is not regarded as a sine qua non, and the main progress will continue to be on a broad front across existing departments.

In effect, the computer policy committee will function as an ad hoc department of medical research computation. Present departments already offer ample scope for graduate stuQy in this field; Stanford also has a well-estab­lished tradition of flexible, ad hoc Ph.D. programs in interdisciplinary studies. However, some interesting possibilities for further strength in specific areas of computer applications in medicine are being stUdiously pursued. The most obvious opportunities are perhaps in statistics and epidemiology and in the inductive logic and linguistics of health sciences.

10

The operating staff of ACME listed in the budget is likewise merely a core group; most of the applications will be engineered and programmed by a much larger staff, in aggregate, in the respective laboratories. Already visible are some 8 - 10 qualified engineers and over a dozen full-time programmers, not to mention the prdblem~oriented involvement of faculty and students in these same functions. This number will undoubtedly grow rapidly; the core staff will not.

(9) Responsibilities of policy committee; preliminar,y guidelines and timetable.

The policy committee, whose members are signatory to this application, will act on behalf of the medical school in setting ACME policy, subject to th~ following guidelines and such revisions as may come from or be negotiated with the NIH as granting agency. Its principal functions will include:

1. Confirmation of the associate director (nominated by the Director, SCC).

2. Fiscal and operating policy. User fees except for more-than-nominal expendable' supplies may be suspended during the earlier stages of ACME's operation. Some expansion of its budget may be financed by prorated user fees payable from current grants. However, our colleagues should have some notice of this in order to adjust their various applications and the details must inevitably be negotiated with NIH.

Initial operating policy will be the maintenance of OS/360, on three shifts with variation of priority schedules to facilitate, respectively, prompt

'telecommunication access; scheduled high data rate experiments; batch proces­sing of larger jobs.

3. Relations with SCC and evolution of the system.

4. Technical conSUltation and indoctrination for colleagues interested in computer application in their own fields.

5. Budget management and hardware selection.

6. Definition of qualified users. Initially lines will be available to each academic department of the medical school. Any research, development or demonstration that contributes to its programs of health research will be encouraged. Since ACME is primarily a research, only secondarily a service, facility, there will be few opportunities for ambiguity that might arise where applications have reached a production stage, e.g. as may be hoped eventUally, for routine clinical care or library retrieval. However, the research and development for such applications would properly qualit,y as a use of ACME.

The committee may also consider a limited number of lines to users elsewhere on campus whose work is closely related to and sponsored by health research activities within the medical school.

1 \

System programming for ACME itself, or in connection to simulating of the /360-67, that may contribute to the medical school's utilization of either machine, would also be a natural and qualified function. Writing, debugging and testing of /360 utility 'programs that will be useful and available for its public library also qualif,y. Lines for these purposes may connect to one or a few stations in the sec and the Department of Computer Science.

The committee will also review and may qualify other suggestions for inter­communication of ACME with other computers, local or not, that in its judgment would result in a net benefit to the medical school research program.

7. The committee will administer the distribution of startup support, funded from this grant, for such purposes as

computer time at SCC

programming and engineering assistance

ACME user fees (as may be set up).

This support is intended mainly to introduce new users, and experienced users will be expected to arrange to finance these charges from their project' grants.

\ MILESTONES

Epoch 1.

The general plans for this epoch are fairly straightforward, and comprise the bulk of the text of this application.

"

October 1, 1965. Submit grant application. Continue technical studies on data acquisition network and "databank". Indoctrination of present potential users on prospect:f.ve facilities and languages of S/360. F.xercises on the /360 simulator and a PL-I interpretor now. available on the 7090.

January 1, 1966. Activate Macy Foundation planning grant. ACME director and engineering support staff appointed. Begin to plan and effect wire connections for data network.

April 1, 1966. Activate NIH grant. Recruit additional staff. Installation of consoles and data links. As interim these can be checked out on the LINC computer; possibly by leased wire interconnection to other /360-50 systems. IBM promises its OS/360 for this month.

June 1, 1966. Install ACME (the/360-50). May 23, 1966 is the scheduled delivery for an 1800. Scheduled system and experimental work can begin immediately. Operate a time-sharing system under OS/360 as expeditiously as possible.

Continue planning for Epoch 2. (User's experience under this system will be highly relevant to scheduling algorithms, file organizations, etc.)

Epoch 2.

February 1967. ? March 1967. ? May 1967.

SCC installs its /360-67 T/S system. Campus service startup. Routine T/S service discontinued on ACME.

Many problems remain to be defined for this period, have not been decided, and only suggestions can be offered at the present state of informa­tion. In any case, there will be a continuous adaptation to the contemporaneous environment.

(a) Extent of physical integration of ACME and SCC hardware. Certainly all peripheral consoles will have optional connection. There is no obvious reason why the SCC hardware should not be located in immediate proximity to ACME, the building-investment being trivial compared to the computers, and the present location of the 7090 having no special technical basis. This would allow a policy choice along a technical continuum, with, of course, the mo~t economical mutuality of access to memory files and other accessories.

(b) Adjustment of user fees for T/S service from SCC. Many users will simply charge these to current operating accounts under their various grants. Additional support might be represented by programming assistance under

,.

) the present grant. In addition, some allocation of funds should be negotiated for medical research users' time on this service (see budget for year II).

(c) Retain the /360-501 This processor mayor may not remain the ideal choice once T/S responsibility is offloaded. For example, the /360-44 could have many advantages if, by then, an adequate operating system were developed, almost necessarily with some hardware upgrading (memory protection). Even within this period we may anticipate another technological cycle, especially in the field of medium size (by then they will be "small") machines for process control.

(d) 24~hou~ serv.ice~ .Some users both in laboratory and clinical research will benefit especially from an opportunity for 24-hour service, uninterrupted by the usual round of maintenance and emergency downtime, or by commitments to other functions. A combination of the SCC-67 and ACME-50 could give an approximation to such service to the point of utility, if not reliability. Future plans for this type of service would benefit from experience on the weakest points that need to be shored up.

Epoch 3.

1968-701 SCC services may have evolved to a point justifying a comprehensive merging of the central processors, e.g., into a dual /360-67 or into a /360-92 level system (or whatever the technology then offers). If our experience up to that time justifies a conclusion that more efficient and economic~l service can be obtained this way, the medical school policy committee has the authority to pool ACME resources with SCC to make such a confluence a possibility. Our general policy is to give all possible support to the university system consonant with responsible administration of the grant as a health research award, and prudent attention to the special requirements of medical research. Indeed, the viability of the SCC campus­wide service is outstanding among these requirements.

Continued experimentation with separable processors remains almost certainly necessary. By then a machine the size of the /360-44 will not be regarded as unreasonable for a "peripheral processor". The main issue is not whether, but which unforeseen challenges will have arisen justifying still higher levels of capability. However, the troops will have a great deal to do in developing the sensors and activators to mesh with the control logic. The development work hence goes beyond the scope of general computa­tional facilities and must be dealt with on its own merits in the course of time.

(11.) Housing the computer.

As an early expedient, the computer and core staff will be housed in an area within the present medical school building which is becoming vacant in the course of some moves into the Clinical Sciences R~search Building about to be completed. This is an area of some 2400 ft. now occupied by the Instrumentation Research Laboratory which will then be a few yards away. Shops and consulting support of the Instrumentation Research Laboratory will be available to ACME. Additional office space is also being worked out. We are, however, seeking means to support more attractive quarters in a small building that might be erected just adjacent to the medical center in an area designated for the next expansion. This would facilitate the eventual

Iv[

~,

, more effective consolidation of resources with other campus computers during the next few years.

Pending an alternative plan for housing the computer, some $30,000 will be needed for remodelling, the existing laboratory, mainly to install false flooring and augment lighting and air conditioning.

(12) Interim support during staff buildup;.:

The Macy Foundation has just announced a planning grant of $80,000 for the year 1966 which will enable us to proceed with the recruitment of a deputy director and engineering staff. (The new Clinical Sciences building was designed to facilitate cabling; common carrier services may be the mainstay in other parts of the medical center). This group will detail the orderly program needed for an effective May 1 start up.

Interim support is also available from the SCC in recruitment, software system design, and negotiations with suppliers.

The Instrumentation Research Laboratory will also furnish interim engineering support. Particularly relevant is a cooperative program with IBM Federal Systems and Research Divisions in connection with their interest in the Automated Biological Laboratory. (See Sec. 5 ). This is not predicated in any way on procurement decisions, IBM D-P Division having had no involvement. The system design of data networks for laboratory instrumen­tation is one of the main tasks of this study which will be proceeding over the next few months.

Professor William Miller (?rimarily responsible for SLAC computation) has had uniquely important experience in real time computer interactions in high energy physics experiments and will be particularly helpful in high data-rate problems.

Ir

.. --COMFUTATION CEN'l'ER

A Pro~osal to Replace the IBM 7320 drum on the Illi~ 7090 ________________ '_o~y. 'a D.E.C:_236 ?l~ ____ k __________ • ______ __

The ComIJU t.ation Center l .. roposes to IJUrchase a Digital Equipment

Corporation type 236 Magnetic Drum Storage Unit to replace the IBM 7320

drum no,o} attached to the IBM 7090 in the Comliutation Center.

Drum Amount of Storage I Transfer Rate ---1--_.' ... - Cost

IBM 7320 about 200,000 36-bi t words

. auuut 34.,000 '-lords Rented at $960 Iter montl; per second (after discount).

DEC 236 20 2 ::1,048.,576 36-bit \Olords

Cancellable or; one [lonth 1 s notice.

157 J 000 'HOraS / sec.. ;~95, 000 -{-- ;~15, 000 interface and installation (no discount,). Total Total ;\)110, 000 ~lus 'tax

The basic I;erformance difference is "i:,bat the D.E.C. dr1.un stores 5.3 times as

much information and transfers information h.7 times as fast between itself'

and the core memory of the cumputer.

Reasons for 1-Junti118 the Increased Performance

1. .!E~e Time-Sharing S~~te~. StClnford is carrying out researcb in

improving interaction between coml~utel's and r;eor1e. Our goal is to make

available to sCientists, engineers and other c!Jrnputer llsers the services as a

I-ublic utility of a higb s:peed computer in liiG office or laboratory. The

aJ.;:f.ilications this vJill make possible and OU1' apIlroacb to tbe l;roblem are

discussed in our r-ro1'Qsal to the' Na.tional Science Foundation for sL1I-port.

The dr m rlays a critical role h1 a time-sharing system and its

:r-e:i:'forrnance determines ho\·; many time ... sharj.ng lwers tbe system can hendle and

ho,\} good a service they 1-11,11 get. Hhen a request for service comes f::com a

user and a' background job is interrurted fOUl" data transfers ordinarily

take l~lace, namely: background core ~ d:cum ./

time-sbarer drum. ~ .. ' core time,,·sharer core -_ •. _-?' d:rum b a c :,~ [!; ~( (;l],nrj, 0::':'.1:1 \. C ()~;:'(:? --"._ .. -.--.-.

In the ""(090 using .tar;es or tbe 1301 disk the four transfers -~ake

12 seconds. ''lith the 7320 dl'Uill they take 4 seconds and \'lith the D.E.C.

drum .9 seconds. This time determines hOvl small a job a user can do

effJ.ciently vIi thout using more time on data transI'ers than in doing the

job. Fl'om e different lloint of' vie,,! drum rerformance de-termines hOl-] many

users can be handled and h01'/ good service they can get.

The M. I. T. time-s~aring system ,-/hen it first got worl~il1g in the

s mmer of 1963 used the 1301 disk file. It Cuu1d handle at most eight users

and gave them rather bad service. The present M.l.T. system uses two 7320

drums operating simultaneously (The cost of following M.l.T. in adding a s

second 7320 ,,,,ould be about :-)226,000 because ~ .. le '\-lould need another 7909

channel and 7631 file control unit. Alternatively, tbe rental '-lould be

$4,81~8 per month. ).. Our proposed system ,-.'ill be a factor of' "i:.'·}O better

than M.l.Tis both in amount of storage and tr6nsfer rate.

l'Then our time-sharing system ,oms plGnned initially the D. Eo C.

drl:m or the eq1..!.i valent '-ISS not ave ilable . Theref'ore, l-1e planned t.o have

only a minimal amount of time-sharing on the 7090 and to c:.;ncentrate on the

PDP-I. 'lllus the 7320 has room for at most t'-lO or three "c,ime-sharers in

addition to the necessary system :programs. Tbe D.E. C. drum 1'Jill );;ermit as

many as ten or even more if a good "Nay can be found t~) use the disk as

backup.

2. System Residence. The present Stanford monitor uses 6 seconds :per

job overbead time reading in systems, etc. \'li th the 7320 t.his can be

reduced to 2 seconds and i-lith the D .. E.C. drum to .. 5 S~COl1d.s.

3. User :r-emporary Storag~. Many programs find the co:ce memory too

small and have to use tal,es.. A tape limited };:rogram may sl-:eed U]! by a_

factor of 12 if' it can use the D"E.C. drun. 'Il1e 7320 dr,m is too small to

allo"\-I allocating a significon"t 9mow1t of' space to user secondary storage.

Financial Considerations

The lB1i.i 7320 costs us ~~960. 00 :per month at a 60;; discount from

the normal rental of' ~t2,400.00. '~le ordered tt in March, 1963, in order to

get unaer the Hire before the 60.:; discount disappeared.

\<Je talked vlith Robert OHen of N.S.F. and he said tbat if' tie al'I,lied

for money to get the drl.:m by May "!e stood a good chance of getting the grant

by September or October. He a.lso said that they expected to avlara. us

;!>2:=:'5,000 on our Ijreviou3 time-sharing }.:.roI,osal) and that this money ,·]..)uld

come in May.

., .• J.-

He :t;ro~ose that Stanford go out on a. limb and orc1er the drvm from

D.E.C. tu arrive in September or October. The lead time VIill probably be seven

months, and j.f we 1'1ai t till September to orc1er 1'le wi 11 lose a lot of "tbe

benei'it of ,it. They can probably delay billing us till we Y.nOH abou-t N.S.F.

\Jle 'will try to get a dj.scount from D. E. C.

~

.:.~

SPBCIFICAT!OLTS F02. DISPL.i\Y 8YSTI:H F02 S':L.A~:::'O?.D ti1!IV!!:RSITY

T1'!io is intclldcd to define tl-;.c ;.:cquii"cncnts fo:.· .g l~ayb:)£tj:d

anG display system to be used by StQnfo~d University

~fuy 17, 1963 JHcC/afg.

1. to 12 j!ocitiorl:J cnch 'tii::h a l:cybo:t~ ... d m.1c.l diDplay.

2. At lonG (: 120 cL.:t::ac !:C:~!J to be c1c·;:c::::-.inetl Ly us '7c11 in ~cvancc 0:2 dcJ.iv~1:Y c£ the eye tC:-.1.

J. i\~::i:'i::y to clicpl.:ty 8:.·:tpl:s ~4i.1d pic;:~';::cc b\.1ilt up f::-on pointe 0:: line c "~::;".:~:1t~.

I: .• C::~:l::3C tc::cty~:;::::

11·':) c::fcc t. 0:: tl:c di::cctly i:11.:0 the CC~~t;tc:: c:1d l:avc cO:::D~tc;: C:l00CCS to display t1:~r.1:

'2::0 ~::.:C!cicc i.:~tL::"~ of: tI-:~ CC;:1:l~ctiO:1 i!] to be clcte::7.lincd. capability i~ ~c~ui~cd.

No !:ypc-ou~

:;. ~::c obil i':.:y tc ::l:! in::~in 120 chc:::.nc tc::s flicl:c::-f::cc and 88G c~:a::~c tc::c IC:;ibly 0:1 c:tci: \2icplay c i:::'Jl tc.:::cG~51y.

1. Th:; l~cybc~l'::d chot:J.c1 hnvc G!:· cha::.:lctc:: ~:cyG i:lcluc1in[3 n et<:!nd.:::d tY:)0":7:'itcr ~:{~:7LQn::t. ':CLc sys;;cn cl:oulc1 ce:1c::.T~.:e m1 ::41tC:::.-..rpt 't:,hcn U l:cy :.~ p::c~occl

~::ld '\ ':"!~::l t:~c ~o:'J~')~:tCi" :;ivec ~ :·c.::;(l i:1Ct::uci.:io!.1, t::'::i.1C!'.l::'t a 6 0:' 7 bit code fc;:..' tl::::: C::~:Lnc'~C:: o-;'1.G a 5 bi.t CC(:c iclc:1'.:ifyi::::; t1::c ::cy~o~i:c1. I::? ::cy3 on c.~iffc:::::-::.t l::c:ii.;o.T:~·cl~ c::c p::cc~cd :::i:'.;:11t~n2o~1::lyi:I:c nyctcn d:ould :.:c.r.~cClbc::­

~11 I:cy~ fo~: i::.~-: to SC rlil1i!:',c,::o:''lds.

2. Tl:c dicI>l~y sys·::CL:1. cl:ould tn::c itD cl:~:"~c;tc::G f:'O;'l tl:c Typo 131 ch~:1ncl of -::1 '.e PD?-l cs::::/u ::e::. ':hi.£; cl:nnncl uill be CCi.1i.1CC ted to ~ L~C,)G \':;o;:~

1. 2.

l~-1;itn-:?c::-,,:"o;:c1 co::e. r.1C:-.l0::Y. ':21::2 2Cr::~I1.'y "Gill bc ad(:;.·Cf.8u ble by the PjJP-l cC::PUl:Ci' t~~ cnn opc~ .. ~tc i~1dc?cnder'l.tly tL:::Oi.tC1: the cho~ncl uh:;n t::c ~~in r:rc::;.C':;:y is cOi·.~l)utinc. The c!~~::ncl C.3n deliver 3G bits evc:-y 10 oic::oscconds n-;: tbc ~cnvc:1icncc of you:- ccvicc. ThUD you uill not nced an c~:.:tc::nnl

nCi.:t:;:~·y. He ~70uld lil~c yo~:: device to :~csa: .. d the chcnncl as ~ OOU1:CC of c~~~acte~s and ccnt~ol cha~.nctc~s.

c:E r.:..:!:~ir.:·_lw ~l ic!~c~-f::cc di~?l.:1y

0:: :::.:::1.:·:-~i:"1 J.c;:;ihlc di:Jpl~y 3. T~c :t~ili~y to ~~o:~~n n~l c~.:1=actc~s t;.. CO:1vcnici.1eC c:': cl:~n::;inG fi::cc1 c1:~:·.:1ct:c:: cet ~ •• ~'!Jili·:'::; .;;~ U~.C ~ li81:'~ PC:l. T~:c Lcv·cloI"Jr.~~nt of -:l cuit.!lble ~c!:~r.1.C tiill

::c'l\.ti::.:c c:J4"!::t.:l::.:I tion. Cne plan is to intc::p::c t Q cc::tain cont1:ol .:::'.:l:.~~~c:: .::lC tt.C:1i:10 011 ~Lc li:;ht pen end m.1otl:c::: ns tU"1:n5.n~ it off. If t1:e l::";:l:t PC:l is on and sec!] 1 i:31:t the cur::ent co-o:-dinatcs of displ<.ly .::::c pt! ~ in .:l rcsie tc:~ :::c.:l(1.3blc by the cor:.p~l:e:: a:1d an intc::,xp t s::':.:r:..:ll :!.~ SC:'l~ to the COr.1;?i;tcr. The li;3ht· pC:1 then ::cn.:inc off t!ntil tl:.~ siC::'.:11 is ac!:noulec1Jcd.

G. Coct 7. D~livc::-y dDte. Ou:· conputcl" uill ar::iva in carly Oc;tcbc:i:, 1~G3.

::~

·f ~

S}?BC!FIC.!\TIOi:TS FOR DISPLAY 8YSTI1:1~I FOR STJ..l·TFORD UHI\8RSITY

Thio is intended to Gcz:r..nc tl'!.c :.:cqui~cncnts fo!." .a !~cyboa~d anu display systcn to be used by Stanfo~d University

I·by Ii", 1963 J1:1cC/ af~.

~ ','

" ... ,

c; • r ~j r : f" ~-:i 0-"

: ,i ~-:I ( ,I

. ~'" :1 r~

~. ':oj .~

r:-.l

'"1 .-' ....

'::

:~

J'" ---t ~

'J

-.~

1.1

• 0

'.'

, I

':-1 o o C)

r. Cl '::.

";J C"J

r.! 'M

r-! r,l

9 .... t)

::1

:~ ,'ol

• :J ()

r: ,:-I

n () .:J ()

'v f:) ~

.8 (.') . :: ['1 (J (J .:J .IJ CJ t)

tj :-.... ;, t)

t1 .:-,: (J (J rl

~~1 C) C'-1 l:-I r-t a :,J ~;") I.) :./ C"j () (I ;>

.-: ·1 I .... : .u C) ..... ", .... v

. ("I

(')

r' .;~ t··f

:.1 o (.')

oI.J 11 'r~ o r-.. r.1 o :1 ~-I

f-' .. ::1

./.J I-I ';-1 :J

,!:l

t)

<'J ~ ! • J

.'J () .,.~

r.. "(J n (j

~1 ~ .. CJ ;-: t')

.... .. ") C:l '4 ~~. Co)

•• -1 .. ,j

o u • u

~~ ~) ~ .~ () "··1 : j ••• , (, j .') r)

< ~

. C"

r-> ,..-. .. - . (:.1 f.' ~.-: s~ ..... ." ,OJ

~J

~ ()

• I..J r1 ('J

r ... ~~ l :: .. (j ~J

:: r-I () P. 0 .\oJ t1 ~ :J 'r' r,r.:J {~ . () 0 '\J o J..J ()

~ g t'l'rj ~-. () I.: ~J t) • I

() CJ o 0 ~.J .~ J .. ~: () (" U , ..•

.~·1 ¥

f-: C,) ~"\ () .!.)

.-f ':.1 V ;.1 0 () (:' • .:J () n :t 0 (,1

·:-1 U·~ r.::;

() r~ ~\ ,;-~ 0 ~ ..... J .:~

r,-J to)

:! r.)

('~ (~ (" r,

,'1 !1 :-.... :1 (j .....,

.!.J (j (J

~ r. o u

,! : (J :~t!'::

~(jU .... rl ~.J "., l:-I

(.') 0 ~ ':of o '-v (J

'.:i ('J (J ,.'.:

'f :1 J.J

. 'V CJ

.~ :-J c.," (j

!-: r~ ';.J I'")

Q Qt')

.'J r. '.-1 o CJ

tl Co) ~ '\ :_: ~.) erl ~J

(J 0 0 ·,-1 ~l (J 0 r': U l>~ ; ~ .r: rj t:·1 ~ •. ') ;: () ('j r:I 0 !~,.

.:.! () .. '.~ (j () r':: Cot (,)

. ~.

CJl If o .:.J o fJ :. l"l " ~J

C) ( )

()

'v r':: o () o :1 ~-I , :1 CJ , ~ "cj ',-1 r-I ~-I .

C,1 t;; ... :: r-I () l'l .:J :; o 0 \l fJ ;. ~ r~ n ~~ " .'J

·U r·1 ::I

() {j (,.J '.-1 r--f (.l

~ ;;:·1 ·M C"1 <J r-·I .!l (.'..04 P t)

orl 'r-! C'~ r,;) roO ,..' l; '(,) .!J c:;

(J :: ..... ,~J r: 0:-1 (l .-1 os-t '1 rJ

t. ,." , ... ! r·C) 'r-t

() ; j ,. (J

(4 "1

~l

t') r.:: o t-••

H ri r -: h r. 1

~j

: ... (') ,!J

• ... 1 t, .. ; ()

f'· !~ r.j :1 CJ ... , \;

r~ Cl

<'J .... J :,

~J 0 r: C) (J 0 0 .et Co) r:i t'l U t') ..:.I () 'r" ~ fJ :: ... ...) ~~ ~J r-., r' ,.1

~ ... Ll l:-I "u

•• .., : I .. -4 1-1 o ::J

h Cj\,:) ., ... , .

"u o ,-, "~l

.:.J Cj C'J 1 ~

(j r'

H t'l r'

(j U r'

t1'!~ r':

.U ·d ri 1'.) .. ' .--' <J

o ,2 ,.",

,~

~.J r.)

i~' ·:-1 ",J

.: J :: () ()

;-J r-·I o r.1

r:': •• ~J ,~~ .!: n ... :,) .\J

'N

: / !1 () 0 .:.1 ',-1 r:' 'J

t) .:-i '() :~ ::-j () r~ :.:

.... '; tJ ~J

I"'~ !: I-I ·d (} :.~ :;;

l;·1 0 .;-1 ~{ , . .1 { •• r.: d C) .:,J

t,J .-1 l)

!.: (J !j n .:J ,.·t n c:; 'N ~1

I' • (j o.:i L'l C) '. , .. \ rJ (j c; ., ... ' rJ (-1 d : 1

,::: t1

"':..i (.l r;j o r.:

(J \:) .:J

Cj rJ 0 () (J (.) (J (.) t) (') OM : / .-1

rv .-1 :.J o

o .. :.: ,;. (.')

j 0;.'

t) ,..'1 CJ ~ l:) P. I-I

·.·1 .:-1 t) C'l () ,.' ,,:

~ . b ,.., .-~

..... ,-I

r1 :-1 ~J ,j .... ) o ()!! t.-) .:J .~.J (';] t)

r;.! ~? ::i , ~~ P ,''\ r.., ,I ,I ,.J ,J

9. (.1

(,) n (.I Cj o .:l a !~,

(J U 0 ,:l ,;l .~-.: <1 ~'\

',J E 1 (') ;: (J ;-1 ~ : 1-; r; ,ot.: () r; o • ,? ",.J t-") .1 () C1 ,> " .... ~ ' ....

~':)

~J .. ': ~-t () ,!J

r' (. (.) (j U (I ~-..

,1".: .:.! :: (J , .:J (J .!.:

':-1 <:.1 ~"", .:.1 ,,~ () r: [-I ,~: \J

:. t:.~ .-1 o OH r-I

';·I·d Cj

. ,-,

t:-I o

t-'l r-I '\:1 • r: (.' ... r.: 0

M (~ .t-f CJ ~ () 1'.. (1 ~ r.1 IJ (,1

~ 0 ~

r.: CJ 0 tJ ,!~: (J :.:

... ~.~ ~J .~ J .. ~:! t1 (J :.: ,. .!.' ·'-1

() t'") r~

o

~ a :, u o ~ ~ ~ ~ 0 o 0 ,-·1 ;.: ,.n ~

(''1 (J r' I 1.::> ('J,.!:! I-i rl t'J

c' ~i ~ (! (:~ r"'l :=")

\',j .-..,

t'l cJ c; to) rJ (-t

C) r~ C'J 0 ;: Q

,::! ':.J I",J .::~ ./.J ... J (j

~ ~ t'l o·f ... ...., (~

Cl rJ :>()r-I o ~~ (J

t'l .:J ',..1

r.! ',J r· '0 c1 r: .,!~ 1-0 G

ri () (J .':"l (J o ':,J ;1 ()

1.:-1 () (" ,·1

C) .n

r"

"':J \.J ("-1

1--1 ••••

r--I ,~

.~ ... CJ .. , ..... ,:J

t) :1 (J .!J

~:: n

r-f 1.):~

o () (J t'; .. /) ::

·:-1 ~J (J :j :; () ~ 0

:I·N ~" ::,." .I~..! r-. ~.I .w C) c ".:1

(.)

:-J

"u ; .. (j

n (J :.,

;~ rot ,-. ~ f..; :>-C.I ,-1 rJ -!J

,0<

~ ~ 0 o '~ .... ()

r~ (J (')

C'.; .:.) (j

.:.1 .-r 'r-I ()

r: (J r:

..'.: r.l

.;..! "U .-1 [~ ~ ()

(J r~ ('. f!

I"J CJ

.:-1 :> CJ

o • o C;) .~ " :> rj o oIJ

".;:) 0 • (J t·; .... " " ....... oW (J

rJ ~'a ~J ,~~ : I rl c.> 9 0.-1

,~ r,'l fJ ;1

() ::J .:.:: 0 (-of :>.

(j :., :., ;-J C'O o ,-,

.-! '.·1 Cl 0 :-J ,..r:: r-: ~.l o [,I . CJ

,.~~ CJ: c t'l ;, (J

• 0 1=' .... rJ :: (J 0 (j (J .:J .'J r;j r: (,') ;J :.; (j :,,,,, C. 0 (J (,1 r i ~ .

:..'" c;

(' 1 .~.J () :: :~j

1--1 ,-I ~.. , t~ r.,

':-1 ~ .. ;:; (:'.,

() .. r.l !', • t'l .n 0.·1 ,':l

g :~ • ......... () E-I .:J r-I

. C'-l

=.f () n ~J

n· !.; o (J

l~·r • 0 t) ~ C) 'N <.' .:,l ~! =1 (J ~~ or-! ': ~ { ...... o () l) :;,.

r: t') 0

-.-1 tJ

:>. 'u <'J r-I ~~ 0 -,.1 :.: r-I ~.J

~J M ::i

("' .... o u

a "tJ ~ ~

l"~ CJ ~ t1

~ ~ (J

• -!.J :;. .... () ~ 0 :.: .. ~= ;.1 r; l) .:J l') ::

r l rJ rJ r) .q 0 ~.

lJ ~ n '0

.~l

·t~!

.... ; ,. .. , '!-I ()

t: () ~ ~

.' t· ·I·t t:~

~ ("j

r-I r. ..

r-I M 'I't ~

. t.:·t 1.;-1

g 1""1 O~-I r.J 0 0 CJ :.1 .:~ ~, 'J OM "0 r.: tIl

o O.:.J f-1 (j 'r~ r.:

t.) 0 o

o r" ,!)

C1 ,:_J 'M

r.1 F: 01"1 ·M r': r':J cj :-J :., .!J.:..IO :1 I CJ (.') 0

:-1 ~4 ~r-I : .. 'I'. It V () r:! ~I >J r": ' o~ l;-f

l~ .. 1'.1 0 C'j

t') 'I.' P r-: o:-t ('j n

IJ ; 1 (J o :,t .~ ''1 .-• i~ 0

Cj (j ~.: ~J fJ ~,:: r) r: 0 ':.J

t<-t ./J ,:~ fJ ()

::l tl U ("j 0

o 0 ./.J : 1 ('1 :~ 0 ~l f) ()

.:J r-4 rJ ::J .!.:: Pol ,!J r.: :1 l"J U ./J C) rJ (J ~J 1.1 r.i .!J"0 ('J ~'\,.::!

c .. rJ r.! ..:::.'.:.1 t<) ;: Co .,-1 (j.u .~

CJ M 0 r" t'l :_1 (J 0 f.1 ,:.J r-I : ~ cJ ? ':'J (J .... .::,!.) CJ ('J .~ () C'J

(j c:; P.. t) (j ,.:.: : 1 r-J t') .... r;,J r ... Q "1-1 .~J ,-I (1 ~

.L: 0 ,::.! (J tl 01"1 ~ ~: U ,'':: j"j {J tl :: . ..., C:l t1 C-f t1 ';01 CJ l.,

(J .:.J

r-I'::': '0 M.-I (J :: CJ rJ ~,

P. (J t:)

H r~. t·) ~ rt. CJ •• J ::1 •

(,.,I'd '1"1 ::: 0;-1 f~ CJ ,~.: ';:} t) ,:-~ (J E~ .:J r: .:-1 Ii 0

() /'j rJ Pv

~ .. r,-J () , :.! () ~r-I

........ t.:l 0 0 :-1 .:-, t J .-1 (J t:"1 r-I

~ ... P.{) t:i t,} n 0 t') 1:1 r: .'.1 r:i .:., .:: ~ • :.: t 1 i.:) r! L') r.:: 'r! 0 o \~ r-I .,../ • I ,. 'J

o rJ t-I o

CJ CJ .-·1

c:i t!:~ ::= t) , I

.;~ ,., r;

tl ~.) 0 .r-J ~ fj

.,-1 0 .:: r--4"cj C1 'Cj

g G '1

:.! J..~ ,.. ~ .. ;J

(J r-t .IJ o 4·~ .tJ 0

t1 ':.l U (J .:J c:1 c.. ;-J ~,

'r! ·:-1 , " ~i ~~ ~; \.i

l:.~ l;'1 \) 0

() (J tJ I:

·,-I·:--! (I) (;)

~() .'J (J

'.-1 ~ l-t (J ·;·1·;--1 ,~) (j \j U

t>

(J ::. . ~ ... 0 E~ 0 . .

o ;1 :j t) CJ ~J

r.: n 0 ~) u

r. ... rl t') () 0.-1 t'l t1 • .:'~ 0

::: \ :)

'.-1 r-I .:-1 ,.:)

-<

l) : I ~ I .:~ Cj (J r-I

r-I (') Cj

0;-1 rJ ~'l (.) t'J I-~ ';"~

..... : d C.1 C', f.:

() -! J :.: 0 ()

',-1 C'j ,:.' ~,

,J

=7 Cj

,r_,

(J --. ('-1

;.: ''0 H '-> i.'J (j

-.-1 •• "! ,.: ~ "v t) ~.J

.-I ("-1 ('1) ........ r.. ')

. ('I) \,.? , .. r-of

... lof CJ

..a o .t.J o o ~

r-I :., ('j C)

r.: .;-1

~ ON f4 :., C'J

r-l r-I ·;·t ... .... 1-4 CJ .u :1 r. .. .... \-' o o :.,

<.1 . (J ~.J

C'J r~

f? CJ :>

,,:J'M tl I-{

o () o (:) . \,;) 1".

-2j:u!oon£ t:961 c L1 A:~H

A41s~eAlun P~oJu~~S aq pOGn oq o~ OO~9AS AULds~p pUU p.:tucqAo:'l E' .7.0; s~'UiJu\)'::;1nbv.z ~\.p Dtq:;ap O=f pOpUO:rU:r, 131= o·PiX,

1. 10 to 12 pocition:.> each lTit:l a l:cybo~j,:d and c1iDj?lny.

2. ,At lc~ct 12n Ch~3~·.nctCj~G to he c.1ete::-=.:tncu by us \~cll in ac.vancc of (:c)_iv~:::y cf the cy~tcw.

L; •• C::~~:nctC:::D t-J·:;;cd on tl:.c !,:cVJo~~:d "'0 cii::cctb ... i~1tO the cor.:-:;t:tc:: ~:1d l:~ve ~ ttl L.J ttl _ ,

':"'l' .... c.":;'.-,,'-"1- 0'-: ~-~'.-. ,·1{('~·1"'-," (·,· .... .,cC"f:' .;.; ... ~ CO-'::1U~-C"" cl"oo"'cn to cl'~"',:"",l<:l'V t1,~~; l. ... '.J _ .... __ v .... ~ .... _ "-.~.'" ... ,,] ..... _"-.1. -.,.,.., -........ .. ••• - - -. ~ _U.J J •• _-

'l.'l:c p:::ccicc i.':'Qt:.1~:C of t!:c co~ncctio!l. is to be detcj:r:linecl. I:!o !:yp~-out ,:cp.:l;;ilit:," :!..c; ::cc:~i:.:cd.

s. 'E!:e abili~y to :J.:lintain 12[: cl:a:::1ctc::s f1icl~c;:-f:-ec and 380 c!:a::aci:e::s lc::;i'bly Oil c~ch clicpluy ci:::ul~[1:1~o~~ly.

1. TI-.c l~cybon::d cho-:..:J.d l:..:1ve Gf:. cl~a::cctc;: ::cyc i~cluclin[; a nta:1.d~::c1 ty:)c,\;L'itcr­::C:lI)(;.:l::(~. Tl:e sys ten nl:ot.:ld sei.!c:,,·~ tc D~ i:1tC:::;'l.lpt \:,hcn a l:cy i~ p::cn sed ~1!1..J ~. ·:~Ci.: t~~e GO:-J~~1tcr 0ivcc a :'cacl i~ct::uctio1.1) t::~nc!":ti;: a 6 0:: 7 bit code f.o:: tl:::: cI:~~_·n.:·~c:: u::d n 5 bit CC(:c iClC:1ti::yir:.:=; t.he l:eybon~c.1. If ::cys 011

c.:i::fc::,~i.1 t ::C:/U03::-C.O ~:::c p:.~ct.oc2 c i.r:;.al tc.lr..cc~l!;ly the aye tC41 zhould l'crX!nbc:­~11 I:cy8 i:o:: ~::1 to .5C r,1illi~.ccol1~s.

2. The c1irpl.:ly aycten choulc1 t~l:c itD chm:nCi:c::$ fZ'CD tl:c Typo 131 cl:cnncl of tLc PD!?-l c(";:-.:~uce::. This cl:n~;,:1el uill be cC:1L1ectcd to c1 !;.C96 l:'O~U 10-bitG-;'J(;::--::·'o;:d co::e. r..c::10:1-T • ';:120 ;:1Cr:lory 'tilll be a(:~::cr.Gable by the PDP-l ccr.:putc:;.· b~1.·:': C[!n Opc1.·cte i41clci?Cl1CC:1tly t11::o~t:3h the cl:~nn(?l 17hcn the I!' .• :::in ncr.;.Q:::Y is c:cr.l;.)utin;:;. The cl:ni::ncl can deliver 36 bitn cve:'y 10 ~ic:'occcondc at the con"\lC~1icncc of you:: c1cvicc. Thu~ you \lil1 not need 3:'1 cxtc::..~al nC~:1~:::y. Ec 'ioald like you:: dcvice to l~c:::;a;:d the ch~nncl as ~ source of c~~=uctc~c nnd ccnt=ol cba~octc~s.

1. 8:'=0 or l:~:~ir:'.:~n flicI::c:::-£:'cc dicploy 2. S:L;c oi ~l:1::ir.!u~'J lc~~iblc disp1.:lY 3. rz::c n:;ilii:y to :>:::0::;::.:1::1 ac'\: cl:.a::actc::s L'r. Co~:~vcnic:.1cc of cl:~n:in:; fi::cd ci:a:::nctc:: !;ct 5. i\'!.;ili:.:y t~ u~c a liCht pen. T11C 11~vel0pl:lOnt of n cuitnblc cchor.1e '\-iill

;:CqUil:C conGul tn ~ion. One plan in to i'i.!tc::p:·c~ a ce:'tui~ cont1:o1 ;';~:.:l:".:1G~C:: ~c t!.1:':1:':10 on tLc li::;ht pen L':aa ai.1ot:hc:: ns tu~ .. n5.nG it off. If 'cbc li.:~:t PC:l. i~ on ~nc1 secc li~ht thc cU!':"'cnt co-ordinates of <.1~!;:!1~y. ~::c ~ut: in .:t rc:istc:' read.3blc by the co=p~tc:' and :!i1. intc::rupt !: 1.:~,:L!. :!.~ CC:.1t to the con?~tc:.~. The lit:;ht· peil then ~c~inn off cntil tl:.c sic;n::l is 3c!:no"\·:r1cdc:cd.

G. Co!:;·;; 7. !)elivc:~y dote. (}I..l:'. conputer ni11 ar:'ive in ea~ly Octobe~, 1~~G3.

SPECIFICATIONS FOR fIN CT3, A NEW VERSION or COLOSSAL TypeWPITE?

The purpose of the colossal typewriter is to allow the convenient

on-line editing i of ~.~ft~X~B~.S texts of various kinds especially

computer programs. Texts may be typed directly into the computer

or they mav be edited from older ~.~KEXk texts coming from paper . '/ .. rt: 1-1- -f-I "\ p,r\i 7C"L".

tape. The output text IS puncher! on paper taC'e~ kl~ &p .... ~

The program has two modes R~ of operation, text mode and control mode .

f

Text mode

When kMKXK •• ~Mk.E CT3 is in text mode characters typed are added

"t4 to the end of the r _5 r with the following four exceptions:

text 1. The backspace causes the last character in the BKffKE to be

deleted.

2. The vertical bar causes the pro~ram to enter control mode.

3. The tabulate causes the program to carriage return and tvpe

JiMII<. the last 20 characters in the _-...:--t., .t (~ ~ h1.. (~f:-

4. The ~~ ~Mete. causes the clich~~whose name is the character text.

typed next to be entered into the BKffKE¥ The four special characters

serve as names of EiK EiiKMKZ one character clich~~ consisting of

themselves. Thus single ,uote followed by a backspace EHBK causes

a backspace to be entered into the EKS text~. Actually, either the ExaEx

cliche itself or merely a reference to :t ~B¥XX will be entered accordinp

to tR~,"lItE.XX.,".u.x the setttb;; of "ilK a swi tild; in the orogram.

Control mode

In control mode t1pKinp a charac~e~ rau~es a certai~ control iKE

operaTion to be executed. Thesp K ogerati0DS are:

,{ A «(,' l M (, r.;.~ttL,t .~, "",-l( C < V" J: .1-~ . ~

,·S: \o,...~ C< 'k.ri. \ j. i' T' ~'l..{ C,) ., (.. (t:" f . ., '- Hi e . .;."

r

1. The backspace KKK causes the last character in the text to be a&i

deleted

2. The vertical bar causes the computer to enter text mode.

3. The tabulate causes the K. program to carriage return and type t

the last 20 the text.

4. The

... name is the character typed next to be inserted in the J P text at

a position indicated bv the . 1<"

pOInter. text

5. t causes the current Indf." to be typed with all references to

cliches spealled out. text

6. 1I CBl,"~' t I,.l ludx"" to te typed with t references to cliches

indicated.

7. d causes the current text to be displayed on the CFT.

8. £ causes the current aKff."xtaxBB text to bex~KZx punched

9. mv< causes the current text to be made into a cliche· with the

name ~ and stored in the clich~ memory.

/ 10 . ko( causes the cliche named '\ to be killed HIIX making the name

available for a new clich~. /

If a K~B space is given instead of a cliche

name the current text is killed.

11..::.0<. causes information to be read from paper tape until the d,t

clich~ named ~ has iust been read. The information is added to the end

of the current text.

12. ~ V where Y is .,niXII:KtaixHHUIIX' a decimal nurrber causes lr'lines

of text to be read crom paper ta~e and added to the current text.

13. ~ causes the cliche ... named ~ to be obeyed. That is, the

program takes control characters frorr. the cliche unt il it is ended.

14. 1 causes the parenthesis level of the current text to be computed. (iu Li SP)

.oj

DRAFT

Request for Support for a

from

COMP1JTATION CEi.'lTER, STANFORD U~l1VERSrrY

Director: Principal

Investigator:

George Eo Forsythe

jolltl ~icCarthy

:Note: The cost estimates in tilis proposal are preliminary and about a year old.

M~y 6) 1963

.'. .. ~

........

"' ....

,~ ..

,~

,': \ ... :.:', ....... ' ... : ..... ' ....

. ;

....

. .....

. :'; .;::::f;.~;::'('·'.': y.:;

... :,.;.~:;::,: ~ : "

.(

~., . ,.

",i,

", .. ,<.

. .~ ; I. ;,~. • .',

~. J

.t'·'"

L ... : , .• ,~ ,.~~.: _ .. I~ :.

.':,

.:

.~

,,'

.' .11."

I i,:.: . R~guest for fltiRRo~t ~o!~a ~aEge Memory Comput~r

I , ..

The Stanford COnlputation Center. is asking for funds to get and i

op~rate, for research purposeh, a cO!lrguter ~'ihose main featu~e is s'

million"'~\lord magnetic core memory. The amount :=equired is estimated at

3,500,000 to 5,000,000,. bpt the project can be stariec:1 ~]ith less.. A

firmer estimate 'H.ill b~ available soon.

This computer lo1il1 be operated as a research tool for certain i ..

areas '\'7here memory limitatiOns are today standii1g in the '(~ay of progress,

notably the areas of

Artificial Iutel1igence

Symbolic manipulation fox physics ahd other theoretical sciences

Those numerical analysis problems that require . a large tllemory

Simulation of adGptive systems and neural models

Info~imation retrieval

and other problems for \·]hich a large J.andom .. access memory is t~quired.

The La-rge }lemo:ry Coffiputer (liID) is not intended for us~ as a general job-shop

facility for ordinary uses; \,,1~ have already an 1B1-2. 7090 and ate setting a

Burroughs 85000 for regular D.~edso A IDp··l computer h~s been acquired for

Sll1~11 ... scnle time"'s'haring. The Ll·fC is to· be used el::clusively for those

resetn:ch pI'ojects 'tvhich need its spee:tal faci11t:t~s, and \\Te plan to make it

possible for it to be used by appropriate zesearch projects outside

Stanford ..

''\'" ." ': ;;:.;.

'I

....

j

'}' "

' . . ,

.......

~.~ .. -0.-....

:. : ; .• +:

.......

~.~ ...... '" .~:

..•. ...

-~ ", lit'"

1

" '.

',:: ,'I .;..,- ......... : ..

.,", .

-::., ::. .... '~

.~.

..... '0

.'

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

. :' ... '

...•

'-. ."0Ip ..

. '!' ' .. ',' *.0' •••

. . . . '

.', ..... ~.;~:~~;~ \ ...... '- ~.. ' ..

........ _ . . '

::.~ .... ' _ ..... , ....... : .

'.,

".',

c ... : _ ~ , '. • .. . .... 0 ...

. :i .

;t., •..••

.~ ... ~.~.

. \ "

'i:; .:. ~~~:: ••. ~2~

':-. ..:.:

.'

,~

. ~. : ' .. ~ . . ·~i~~ :-: ',-.. :"" :: .-;1

.... ,. ... ' .• 't. .. -....... :

l ,~-. ... ~ -,. .. .... : .... ~-

.~

~ •• 0' .

,':' _. ' .. :':.f: .; ......

,: .... , ... :. " ..

.. " 7.,

:..

','

.. ',"'~ ......

- :... -, ;":; " .. i ~~: . ::-" .. :.~-..~.. ...,' .. ~: ';. .1;: ,,' .~ .~ ',i t:;.

.. \ .~ . , •• ~ ',0'_

"

" \:.

-;': . ;~':.' .. .......... ~-:'." .' ':-::: '" .:: ... ,; ....

:. ". ~ .-~~' ~ '.: .. : ~~ \~.:~.::::; ~.,,:

.. J . -':.,

~ ....

"

:. t

'j !

'. I:: .

,;

For many years (since 1958) the largest gen~xal1y available

computers have had 32,768 v70rds of directly addressable memory. This I

limitation has long been felt. 'J:oday, usin.g modern automatic programming

techniques, it is easy for s single prog~mmru:T to build a system that

e:Khausts this m.uch meillOt'y.

Very ~~ch lar.ger ~2~Dries, directly and itr.mediately accessible, are

feasible today. They are required f04 the progr.ess of computer science. The

reasons for this can be stated very simply and generally.

The frontier in computation is c01!"tple::rity. To a large elttent, "(·1hat

computers ,·li11 do five or ten years from nO~·7 'Hill be done through greater

cOO1plex:tty of program structures.. 'J?he large programs that do net.] things will

combine features of many things done today in sepa;:ate programs, together ~)+th , i

further innovations and methods. Not. only ,(-7il1 these programs be very large

(m:td require immediate memory access) but they '''7i11 requ:L:i:e access to veri l~I'ge

amounts of: ifPWp d1ately acc'eGsible data representing, so to speak~ the pro~raih' s

This argument :A.S Sttppo:rted by con.siderably specif:tc e}rperience41 I1t

~

the !-Io!~ To P:1:ti.fic.ial Intelligence Proj ect fI J" R" Slagle ~n:ote a program to do

analytic integration using the methods taught t1.1. calculus courses 0 The

program \'j.;1S successful:. but '\-10S bsz-cly squeezed il.1.to the 7090 memory. As a

resul \: se'l,reral techlliques had to be left out alld it '(·TOS n~t possible to study

~-:r:hat thai:: effects ,·:auld be.. A c01'!siderabl,(! part o£ the effort ·\-7.;'18 diverted to

the unprof:U.:able fight against the space limitations. And tIle more amb:1.tious

plan to embed the integj:ation p1:ogram in ,'3 lcn:.-gei: package for doing analytic

CC11cu.:tC1tion.s had to be shelved ..

The same fate was met by progr~m3 for proving theorems in predicste

proof-checking

pJ:occdurer::, and simulation. of ttlG IT:.ental processes o~ C'l chess player I>

, . . ,

11')~ t

r .. -,

:.- C v .. 11- i

1

--1, ,u I

be 1].. 1.

n '~' ...

.;, . ,. -. -I.. '1.< u.

.·.fo'

, '

r n.!!: ..... ll: '1: .•

J.

L

, .. r

(

r

t,

)1'

o

, , , • •

"

t ,

1

o

"" '.0

( ,

. - i· ... C ... ' •• ,I "

(n ;., ,. ,1 ..... i,

: Ll c.r!

•

t:' • ~ 1,....-

1y ~o

. , L )

.,

)'-e

I '~

St t~.

. , C • lao

. ~ )

, ,

v:

1. t- '" -1

'. 'J

-,

,.

...1:- -.'

, j

r. ... ,

.'

J,Kl" l'l~

, C'

32

, "

'J

c .

1 c

';L./5

( . , ~

30

l

J

/' J < . , , > ••

, 1. ..

r· •

~ . ,,1) c.:r

P.

"

:-; f:

.,--

. ;;i.e. ("'j")\I:!

hc.r.· r~ -.:~l·~ . ..,)l ~

&.. ,.' i. f]:

:Cl. ::;If,:'!-,

c:· {

r 1 t ..

[ L. 65 536

, (' ;-;

< , J. "" .....

I r ~

·.\

.' ..

The of I .

speed. As

t ., ..... e" • II. 1"'''' • 1 ..... ~,~ '-." !'1f"!.;-:. :-.•.. ~.: ......... :.'_" ... '_.".~ . 'L~G ·t..U:::l:!,(t :t:::. com.u...!~ C:'.Cl. 1 ... ·Jl .. ~Ul....:;2· __ !;: _::'" :.;

J

~:,"~r.d to limit take hold:

is economically ·J:e.nponab1.e ... Then only' Q

:ts

2. Dcci.gl1C:CG like to assig:.~ special :;:Ul1 . .::~tioi.'lB to as r~an;r.

s~~ces fo~ addresses. "

sl Y'!l1BU the cost ox memory 1~H:C14 comc:s tlO~·i11., lCt=geT mell'tOl:ies can be

'H~ ... 1. ,..,·l1 ._~. ~u. <;L ••• ~,

: .. " .• -\.;--_·.~_.htU·~',.,;:.~ '''_'\~1.; .• -3. -; .... ~_.;~,(."1.~ t·o .. ~t:"t·!··""-"·l .. - .. ~~"..· .. • ..... M .... _ ~ t! > .... ,~ ... L~l.\;l:..1. P .... .-.'(.'1: •• ::,;;v.

Recently,

that

...... , :'0' .-. .... ~ •• : c· ''111 0.4 ... • .... _n ~il . "1." P_,,_ b ... ·i..r._·c ... ,.;:~ _·r--J."Om •• elL!. ... \ L"'·=~I-:" ..••. ) __ .y ./ t.l~.... _ __

, .

The di.sc:o:epan.cy.:Ls tYl?ical1y that bei:\·jeell a core-memory accass ... time

6£ 5 ~:tcrosaconds bnd a dtunl-t'otaticn-tin:-e of 17 ,000 m1.c::oseconds~ TI1US, for·

tl1.ose calculntiono that really t'cquii:e j;'an.doi.U C1cceDO .'the slcx'1do't~n facto~ \07111

be ::. matter o~ httn.ch:ads .. There is, no general .... ·jay to gat around this"

{~:tmublt:, there is 17.0 t·~C1y to make a .1ist ... processing systeu1 l:tke LISP or !PL that ,

~}ill not reduce process speed on large p1.~og!:"mm:; to tIl<=! ralldorc.-acbesc speed of

1n ccrtail1 'cases, pai.:'t:tcu.l,m: proc~sses ccn be p1:'og!'mnmed to trtalte

nut 011e mt1S t not conclude that, by sufficiellt

i113':;1.1t.15.ty; .!?E.Y. pt·OC.CDS cal'). be so t"G~:n:ral1g\:!d! ,. ' FUi:' th~rmo~e ~ even o;;lheri tt~is; CCll.l be

I

dOl1.c, it ,

adds the com.ple:1tity of m::uag1.ng t116 cto:ragc hie::arc:hy to tlie colrlple:g:ity

of tihe problem being sol tTcdti Il.lc" . .r:i.tublY, 'i;h:i .. s l.imi.1:0 the cor:lplc~~:f.ty of the

p!:'oblcns ·i.ihat c~n b~ solved at a given stage of: progi:.'C'lli.iliung . and ·:::orJpute~ i:echnology~

1 .::~~ "o .. ·t" t' . ~. L ITo ~n e:!~tertt ~ one. can. chal'ac~:6:rizc t le .~.1.:IF' I.. c.:.U cla.ss o:c problelJ1s that

O~C can h~ndle, in a 2-1evel ~1stem)

But th5.D docs

flhi.o :J.s thG C.!l(;C fo;: thane syst·Z:lnS 1;·]h:tch are

., ... • • .. to:" conctcntl;t mOi.1.itol'in~ its

-.'_" :~. ·l··.-.·,~,.r:>·:.." -"I::l""_~'!! ~".~ '_I. :1_·1."'-t.'_':!.~,·!.f!t/,:1.· ':."!.!.::: C .... I: .. !..~'-•• _"O· -,/1, 'C"'O .... tltO·- -.-.,...., (' -." "'--"('"' '-:t' - - - l:O'1O ·10''' .... - - - '--- .... '- .!. - - I •• ~ .;, ... !._'-" \ '"' L:.~).:.~~\Jt,;.,.;o >.JV.<.J:. J'-*l •• :.;.;::., .L 1.. .... 5

But ~'7C claiI.ll 'that tlie

the mujor cost 0';: tlle

At that time He hrid a pl'etty goed a~su:::-ance' that 't'YG could 1:my rnemol.;; at bet4~eel'l i

$0'41 O.B to $0. 10" pel" bit in d10 quor!.t::U:i.·~o,· d1.scusscd;. but $0" 95 \·jas the moat

p:::obsb19 esti!i.!ai:a. In each case ~ c011vention.nl cO:!:'e ~lcm.ory in tl-i2 5-micrqsecpn~ ,/ .. '

speed range \;·ias ;u1'lderc"Cood) ~ssemhled l.n cO~1plete Ul1its uith 'bt\ff':!j".~ .r.eg:i.stcrsl

equally but !} .

Good mau""l:U.tchine. intc:tactj.oil~ , i

e;.~per!.sive to meet~

m.~n"'m:;1ch:Li.?i~ iD.:tel:'llctiol.l is t"equi!"(:d to debug them in ,~ 1"GQsoi1C!ble t)~. I'

Hd:i.4 ebvcJ; !i .. ;.

scme of the rese.i.1!'ch 't'7o:t"k reqnir:b.1g lCJ:ge ttenrol:Y aloo 1:'oqulz-GS 11lan-ll'!i!ch:i.rte

l:.e:r:t fI

cent~nl s6op~ 6ousoles.

2. U8G~ fil~ facilitiec~ 6.g. dis~ ~iles.

~ .J.

I,:r._f_.~l:.i.~- .. ~~.-!...·.t-1. o·'~-:"'_'.1{::;' ';_"'~ -\'·-J·.:_O"!!l"_' •• -1.10 ;-1:-;·:1.4\~~, .... -' ·\···J·."'-·trl. ...... : • .: .... ""j.,,..,..~ .... ........... '( ,.. ... _,···,"",.:c-- - - -- ~..... - .. ~:.; ~ _ ~ ~!....- _ ... t._.!'..::.' .. !.. )v~·.':.!. 1.r., ... :c> ....... u6 - Ij.

compili:ersv'

1. He dcr;:i.g::l LZ·:LC· at St:c;xto:i:d ill a vCJ:y Sh01:t time",_

'to l~ei.h.tcG -design 'i::Lmc ..

• -7··

3. L1f:J should be btt:i.l t out of t

~ ·s'i.:ritic1ni.--d l:Lnc of: plug .. in circuitry

CDC. A oyotem XO'!." v1hich design ~'lf.to~tion programs al~e available

i'lOule! be prefcrL'ed espec1.nlly:

flU to~:.atic!111y '*.

Lh He try to get a large company to uire it for uS.o

5" \'Je use a cO'i11ptlter like the 1?D.l?-l i;o bandle all il'lp~t-Ot1tPtlt in

r.m.tputa

order to ISvoid the time-co'nDumin~ t·ask of designing a lot o~ input-

output .circui~ry fo:: the m~d.i1 !"fu.1.ch:i.ne.

1£ thiD pl:'og1:'i:lnl can be fol1o\1e~ the 'computer can be -ready in a yea3:,

co-ope:~ot:i.c11. i .. 7~ tl;

.. I .. tlJ ....... I.··.,_~o·,...cn.~o' ~..:." /,'j1' 1 1 ~~"'I·'·;-"L".I . t·"I a \ ~ .., ';.; \.,~ :Lma S:.1a:::GC ,'r .. !JJ.t;-. \:J:.:cq (h:uffi, t~lpes, disk, t)L'1~;:;.~:c!', J.2-console ~,c6pe-cl.isp1ay OYStc3!:i)

ll'OO,OOO

600,000

500, OOO~: ............. ~~ ..... ~

Total

Opc::·.atiug cos ts pe;.~ yet:;::' ::i50 /)00 to ~;200) 000 ~':.:ri[;:;:-c:!r~7d:tU[~ G;y·St·21~S 1·7(i:;.:'1:: :ts d':::r;.c:! J

1!,500~OOO

200,000

1~ 'iOO ,OOO~'-I

labotatori~s

-8 ..

Because the central processor is the cheapest part of the sy~telh it

can be replaced bya more advanced processor later. The initial proce~sor must

be kept simple to reduce deveiopment time. The possibilities are:

1. An echTanced instruction system using as much implicit address:i.tt..g

as possible to reduce the number of memory cycles needed for

instruction reference.

2. A multi-processor or sL~ulated multi-processor as described in

[1 J can increase speed by a factor of up to 10.

3" As engineering developme11ts reduce the cost of memory more can

be added.

lr. R~al1y good display a11d :tnteractiol1 systems can be developed.

These expansion possibilities provide a future for the engineering group

recruited to 't"lork on the project.

The construction of the computer system is to take one year. Computer

In~ojects in trnivcn:sities have often taken so many years that the project loses

interest before it is completed. The LHC project has a character quite

different from the usual itnext-generationU machine project, and this makes it

possible for things to proceed much mC;1~e quickly. The short development time is

possible because:

].. The design plal'lned is simple. "tsie propose to restrict sharply the

inco:::poration of radically ne"t"7 computer ideas and build a single-adch:ess binll1:y

computer \·1ith a si.mple order-code) not loaded ~-Jith l1featuresH• The design of

such a machine io ,veIl undeZ"stood.

Reference:

Ju~lcCnrthy - ~ime-sharing computer systems, in Management and the Compu.ter of the Future, :i:f.loTo Pi.:ess and Hiley, 1962.

• • •

.'

we '.til! i;;;..C! ..;o::_:. ... r:.::!.211:t Dv,:1.~l.:;.:.:'lc 0';;: - i:'i-:'::-F!·~-=.!.:: CG:';"_.IO!.!~!1i: cal"c!~ •

p.c'!"::.:lrC:l en CC:-.1PO:t2:I'lt:J 01:' c:V:c'.1itc i~ n('~ :!.1.~-... -c::''':cI,) . '.;:~~e cOr.lpc:1c:,'2.:':'C, ,::psEnbJ.~T

o!1"'t~ de:: · ·.3~ !:.2thoc1s ':,i11 i)c tl Gtnntlm:d :Jy:::tcr.! ct:ch ~~ ::'!:.:t of DEC or ~~BH; since

~-:-c do n:·t need unuct!.t!:":. ~p~c{.~G or 7.'tOCCS of op;.,.:;:ctj~Ol1 .• If ~hc:: =:!lcctcd !~2r(1~:<J~c

cYGtem i :1 one f.:Jmil::a~:- to i:h·~ ciec:;.cncr, thil1G~ Ghoulcl proceed very quickly .

3 . 'J.'he specificQt5.ol'!.s, e.g. 1;.5_~d of oi'cl:~:: ':ode, are alrc.::ld:1~ "';-:-01.1

uud':!I"st I: 6 .. Host ifl!!)Ol'taat, t.lie p.t.cI::'~Z~d PDP"l :.nput-outp:!t =Y:3tem't-lill take

c·:U·C of \I hat hos l.11ucyc bocn the :-':»Gt CO!~l::c~~ed pnrt of compui:er design .

'!'he LEe, if built :oon, ~'lill bo OJ ut1iqu+'} facility . It '\;'11.11 permit:

i."C, ~.::r.C!1. :hnt canr:ot oe dO .... 1C c~Iler.~icf1 ar:c1 'oi.!.'i ilclp 'Zi.lOJ_ntni'il OUT cOtJi.;,l.:rJ-' G

Iced in :1 mpt:.ter ;:'mrk.

Ph~se 1.

Gi"'}'cn stl'ong Ci"!,~CU1:~Z:Qr:'C::'T;: ~:e CDl1 Qo::ci:l.:::p

Getting dccigne1. .::nc1 ?c:,:';ui.!.!l::!l

.?t'c:?,nrin5 forr.'..::l l.'ro)~['.:1~10

'r2cz:Lng ::11: or''::e'~ cot..:c

. :::lliG-:':i113 t: c coop3:"'~8.t5_on of 0 ::.u1::,~'...~:(cG·;;t1.rc~ (J.:.:d :~1."'(;.e?,:i_ng

the 1):~1c of c:C'cGuii.:ry.

-10-M- DAY

----------- - _ ... -.. --- ._-¥--¥---

--- -----------_ .. P------

\ ------ - - - ---_.-i.u::../:Ll iai.-:Y Design (

',--Des i.gn

Auton='tion -- { ___ .i_/ _____ ____ .

Del:Lvel:Y of: Compnter l:rit.c S 0 t!lC racOO3:Y

Debugr;ing Hn ~!:, inc

Debugg c.ng System

.'

Dc).ivc:;:y of l'DP- l inpu,'i.: ­oui;;?':.~t CC2puter

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

» )

!

-----1'---------------~ I

-----------_.

--.----

-;tc be~ievc: ~"::1a :; '::lio ::,.chellt!~.e io :;.:e~].::'ct ic :'.i2 ~!~O a~untG 0:: t ::.:It2 ~8si3i.1cd in e .ch ct~p .:lr .. 0. to t1:0 -lc:r;::cc 0:2 c -.-o:L:lap ~c::':::'::.zsib:c. f.l:hC"i:,e ":::.11 certainly b~ ,,- :~~.:: s :!.::"P!J.:!Gc, Di.~:: ~;. .... ~ ~r~nno~-: ;;c Si.~:':·C il~ \'1h:l:~h =: tc~) !:hs 31 :i.P1"nlgc t;:-ill tnl::c place.

T

H

s

(

\ .' 1

A"I o"'nt fU:" e1.ti...Ld1r.:i, C Imput I:) witL larce ::It:llo .. i es---e.& .. _ c'L' • I{)( f'l - c" .. :::E

stol'ttle.

Wi thin a ve,y 1 '!\J years. ft r reasons o' tJ ined oelow. it" L. oe

usual '01' a llliJlion-dJhllar computer to have a very lar&e memory.

The more research-oriented or&anizations oU&ht to anticipate this

d"YeJopment oy ,ettinc; the i r's earlier, even at some;rhat crea"ter cost.

S0 that the industry at larce will oe lIetter prepared to use them. In

particular, M. I.T. OUCht to 'e amon, the first to explore t his area---

namely the use of this kind of system, oecause of our responsitility

tv do advanced research and to prepare J~;/ our ,raduates for o~tstandinc

command of the current tec hnololY '

2. The oUl";;sin arpent .

We have .een advocatinc a million-word memory for two or three years.

wten we aecan this campaicn, the proposal seemed to involve ,reat cost.

Thus. at current IBM cost, the list price of a 22u word 48- oi t memory

would seem to oe around $48,000,000 .~cause of the sales price of

a.out $1. 00 per ,it . We have learned since that this is an art~eiciallY

inflated price, and an extremelY inflated one at tHat. In larce quantities

the indust,"y sllPply price of memory in work in;; adGlressatle ,oxes is now

around $0.05 per lit, so that this lar,e system should add somethinc

more like $2,50(, .. 000 to the machine cost. If we talk aoout a ##1

larce machine with order-c ode capaoility and speed somewruat lIetter than

the 7090, we micht oe considerin, a system cost ina: atound $5,vOO,OOO.

The lar,e memory would then seem a reasonaolrtl .alanced ~hare of' the

are ~lacin, system. We put 4:tls arcument "arly in the exposition so "th"t t·." reaue,'

,<ill npprecia te that we are nC' lon,er ccnsiderinc a radical eXJ;ensi ve

expe':'irnt2nt, au'C rathel" a reaslinaaly proportioned one. IHcit.:~a. ,. 5 L I:i '&u,ments

oppos1n(, t.he 1& r,e random-acct'ss melMllllY evaporat.e when one consi ue:s t),e

¢¢f more realistic cost scale. -----:..;::.:~

ill? r rIlliinder of the e~sn) is thus 'eab<:d on the feet that the JBrC;e memory In

.'..Ie~t~·.)U ~'e.t'reGe!lt;3 an In.vestme!l of the v. a.er of ne ur twu m11.lion dollars

in a.ict i tlon to Ii "cmpute . s,Yste;n costin&; 1lJ"1Y"Here from 1/2 to ) million do Uars .

'rnc 1). ~ c:>st is oased on th", u:-oposal of I'iCCart'1y to study lar,e ;n"lllO' y in

a<eoc'>!.t1 'n with a fast !l'Jd po""rfu.!., out not very fa'1cy, central processor.

T .• 1 'e million estimate is .ased OJ. the proposal to associate the lar,e me!llOry

~ ith e ve ry fallcy. extensive system; the cost could ,0 hi,her, out Drooaely would.:)' t

for a .,achine to oe deli verea. In, say, ti~/ 2-1/2 years.

3. Memory and the complex.!. ty frontier.

AlthoUCI1 many "experts" enjoy denyin& it, the conventional iDlUimcuu

size of 32,000 ~ords of primary storace is today a serious limitation ~I

for new ima,inative pro,rammin, experiments. It is pOBsitle, and indeed,

rather easy for a doctoral stuaent to cenerate this much live pro,ram

in the course of his thesis \{ork, especially if he uses one 0 t!le

modern informRtion-processinc lBn(U8.,eB. (If he does not he is already

seriously handicapped in other ... ays.) Wben he oe,ins to reach capacity

lim! tat ions he bas only one recourse; to ,0 out onto s~condary mem:,r,.

This invulves tllKlm&l.mBila~ three costs:

1. The coet in machjne-time of access to secondary memory; lo"din,; and unlOl>din,.. locatinc files, etc.

2. Tbe cost. of the space ... i thin the p t'o,re.m 1'(,r the second"ry file pro,rrumnin, system itself.

3. The cost or the aed.!. tj onal research and p:'o,rBJ:llllnn, com;) .exi t of' t#/ "orkinc out the sccondary system.

Tt,e tciru cost 1S JSueJ.ly ilIlcred .y ad,e,cates 0, indin, efficient . ye

to ma~~ toe .est oP s~ll memories, .ut I feel it 1~ the most important,

in nW".J ~9. • t ". ..,;.! , u.dnit1onal imolwctua • 'cad or. ttl p ),r~ ... c:.

and .11·_~Y tur!~ Oyt f,#/ t I>e the 'P'JB.jor 1', co·t i1. researilt . T.1is u""i It oS

."':;.. - d e.!.'f' j .~.l Y lID..iS t I> • paid 1'01 n a rel'i" c <- lEye ... r '1.np t

c CE'}lt a ~P! " of r~. pi e.,. "

-3-

Old-timers may have difficulty in understanding ",nat we mean

oy comp exity. It is a.solutely true that in a very few instructions

one may oe alole to construct a process that is cOiDpikex---even mind

aOl&ljn~. The effort that vent into pro&ramming for the IBM 650. with

its 2,000 vords or' store&e vas very impressive, and some remarka.le 1;',ings

were "queezed into this space. But ~~~/ it is oavious that not evt'y

process can ae so compacted. We are entering the era of resourcefU~

programs that have many parts--so many that the program Inust contair,

some sort of information-retrieval system for its own use. ~'or these

complex proalems the task of ana.qsin& the pattel'ns of su.routine

usa&e may ae hopeless~.q complicated. If one cannot make such an analysis

then the use of secondary stora&e may entail a slow-down factor of

thousands or more.

This difficulty is particularly apparent in the research and

de.U&&i~ phase. It is proaady &enera~ true that a system .fiich

occupies, say, 32,000 words in development, could later ae reduced to,

say, one- third that sixe, provided that the program is not mostly

advanced list-processing material, or lndispensiale random ~¢~i~/

access arrays and ta'les . But the reduction would usually involve

grwat and uncreative effort and a len&thy period of time . FnffinmAnmmnm

Even so, this means that the larger programs, which when optimized

occupy a larger share of the 32, 000 :nemory, are not reasonaltle candidates

for development at all.

'+. The argumt:nt a,ainst proponents of ef1'lcient,a'..ltomatic,seconda,),­stora&e-wana&ement syst.em·'

'l'here is r. presslrlC need for automatic systems fer mar,a&e'n •. nl ."

pro It .;os. A B 'e 81:'- J. l t 1:hJ ('" .....

" , ur

'LOSp pl-esence in m~mory is required only in discrete .~ocks of time

Of~UI"'ic:ient lnncth and separation. lIhile this is often the case

ir: p"o!';rams whlcb operate in lenctby "phases", first performin,,; one expensive

tra.nsot'nnation of some data 'lnd later performing another leng-thy

transformation of the result, it is I,ot the case for moce complex

scphisticated syste:ns. In particular it is not the case for proc;re.rns

which, tho~ organized into clearly defined "hierarchies"of

sUlroutines, operate in a vertical mode, app~y1ng the whale structure to

the data a little at a time. Tbd:s is the case in systems which are

"rich in decisi,ons"n-i:l which the system ~p!l!Dri!JBnlluwibnillj!mil1lm!m monitors

its ow[, operation, rather than relil1'l.ulshinc control for lona: perioue to

autonomous suoroutines. Now 1 claim that the decision=rich systems

are, in general,. much more powerful than t'pi/ / those that"delegate'

most of the work--- just as seq'J.ential decision methods always dominate

fixed·, samplin~ methods.

~J

5, The ar~lment that complex list - processing systems represent the work of

restricted 11 special interests " .

It seems to '8e a fact that the stron,,;est advocates of lar!e memory

are,today, these who work weth list-processing systems---the so-called

"Information Processing Langue.,es", It is also e. fact that the workers

in this area are primarily concerned with complex pro_Iem-solving ptP~tr~/

processes---either under the title of "artificial intelli,;ence","simulation

of human thoUl;ht", or ather such grandiose titles. One should not oe

~ misled "y the :'act that this is a small, thoU&h 'local,

group. Thetr real concern is with the frontier of constructing computer

pragrruns to do really complicated pro_lem-solving tasks. As such, their

wor~ forntS, I think, a prototype of the sophisticated computer programs

of the relatively near future. As these workers venture i:;,;o COI:1pIE'x

prof,rams with new ,--:1ndH 0:' or(,anizations, they 9,1'e f~.=l1n/( the squ,,'eZ8

"

('!.t~ "l:l't 'r1'-'t''[J,~y Sl.['~;. A. (;1-#// VI ~:::: •. 01 • iTI ::tLe r al'p.!l<" !IlC··ie - n · ... 0

t' ,In :JOlT (;Orr.U ,O::f. 'p.:(~·,r~l: ,_:t' th,: Lr owr i'_a., it 8' .;,s .'1 " t

, ~i i't t" .., t (. ,:1. .. _ e~'lC( ':nt~' : :rt.: 1, y- .L. i.'l'i.c. ,._.j.' ... ~e. ~u.t"r . "

~.:11e Wr):·.KC:Tc t.0 X";'J 10 'r.:~ neW' o-gn ization:-! . a...w:.vst :''' ... !' t. tJ.€ i r ,"\on ~aKf',

• ~.ll enelJunte:' tJ;e ~eFtrlc~i0 f: 't'~t. i.',11t.1 rnn.'inta·':J that tIlt"

djff"j,c;jJtle~~ f.>rl~ ',/ ch tl P she~r complexity 0 ... t:.he ,lJI'ogra!:ls) rat-h, .... "" tha.n

'litn the na LtU~C, 0!' the SUit .1l:ct. prn"J.ems dealt · .... i t4, t)y the~,e pet,;!"J

Thus ill two irlstance::; j.n my area ---t.he calculus pro~rarn 0 S.l.a.gle; and

the "Shet..chpad H S1i.~tem of SuthcT"land---the pro~:qm:: ~1'llir.mmnp..1a, URe Uf)

.:ll(,st 01 a con'lenti')cs.l sized ;nemory an1 are Tunnj,n~ into diffic.ulties

-'ecnuse of' thir.. J:lut inspection 01' these "yr.-terns will reveal that tlJcse

wo::-kers have not even .egun tc :neke use 01 la"ge [,[les of "-factual

-n.lonoatJ.on" for thei:- proa;rams: tl:r. size is still mostly pro~ram itself.

The trou.le is due to ol'~nir;atjon, and not due (yet) to the addi tilJD

of data which in itself has tc oe random-access, In fact, "e are not

nOH realJ::1 prepared to encourage mllch researcr, into the crt tical proolem

ot· progra:ns "hose operation depends continuously un availaloili ty of

lc.rgc amounts of "declarative". or "cormnon-·sense" factual info.nnation .

LarfOe memoc'Y and tl-,e-sharil1li' in the near futu-=e, the stand!ird mode of

It 1s our thesis that/fmbnrrmnmompm1rfttlnrmlJ'BilrnneTlllilDl.ilJnil

operation of la ... ·!e CO!~4lut,,-r sys-+~ems 'rJilJ ....,e throl.1~l) e. net or' simuJ..taneou.sly

-'::,pe '_ a t.P-a. rePlOt.' C 0:-. scI c: S 7!aC! .. wr,;',ct .r~ive!: Ids 11se!~ + ill lS~~.(':!I of

'rd.i v ~d~.ta... .'{'''£ ~~ r,t..- P' ., 1 TC,f.:. H.acY·i'1e. He ",'1';"1 nr .t n i ~.erc. tE: t'J':r ....

a l'~urr 11\',S ...... ~. ir fr:: .t.'l,p~lt '0 tpi liTj(~ or' 'rac.:.i,I]' O""""'ign 11.10. (,. _1.->

,.. '1,:'S Of I' :',E ,?, t~ l"f:;"i ti~ :c'I, L:L l. J. :~

"".'" .r· >! J L .,

-6-

.ituations it would have the effect of giving each uf a Ulun."r of 'lSens

access only i~ to asmall share of 1;he exjsting lar!,;", "emor", ).lore importsn"t

t'or tlme-sharing is the notion of the lUore-c·r-less permanent re~ide:,ce ¢1

in memory of many frequentlY-clsed liorary routines and systems. If fully

developed, tHis approach will yield, proeaoly. a new ~.p.pf¢/ I technique

of programming and program organization, for a crnnpm .. amm cOlllplex

user program may ~e represented in hi,h~speeQ memory .~ a relatively

small incremental _loc of pro&rrun. Ideally, when tne Itp~1 live

lierary is well developed, the user programs will need to represent portions

not very much more than the executive flow chart of the upper pff>#f># economical

of their program organization. It vill ~ecome ~fif>~¢f4#i'-, in particular,

for programs to make much more use of internal sym olic representations

and languages, since the translators for a num'er of languages call

stay in Ii ve memory and seTve roo.ny user programs simul&&neously ,

In pal"t~cu.lar, the costs 0:' aSBem~l.ing and compiling, etc,. will

.ecome extremely lov with translators desi,ned to remain in live storage.

Wi th tilis, and perr.aps a fev hard"are l'efir,ements, omenoClllllr~mI!.DT!ltlIn

t .,ecomes reasona.wle to JIBUltSlnmmn run prcgr~ms

tha1; a re in y", ry neal'ly a sym. o.cic p:rocedural- lang'J.!l.f''' ~'orm w'ttl ,,11 tt:e

a ttendl'; nt ajvurr'.:;a~es of i'lexi .111 ty apa so· :ree-.lant;ua,.;e ie.I~F,i.n~

.!..Il.:.'e . 10("" 5, J .... ('~) r 1:;;. l t wi) sl.il.l pl.:i:" t

\

" -i ". .

.. '\' \< > " , \ ,