00100						August 31, 1970
00200	
00300	To: Deans, Department Heads, and Principal Investigators
00400	
00500	From: John McCarthy
00600	
00700	Subject: Computing at Stanford - Long Range Plan Project
00800	
00900		This is a reaction to Professor Miller's memorandum of August
01000	25  of  the  same  title.        I  am sending the memo to this wider
01100	audience rather than simply to the Project Director, the  Computation
01200	Center,  or the University Committee on Computer Facilities because I
01300	think  these  organizations  have  bungled  in   the   past   through
01400	technological  laziness  and a leave-it-to-IBM attitude and are quite
01500	likely to do so again.
01600	
01700		My main complaint with Stanford's computer policy is that  it
01800	has  not  made  a  decisive  switch  to  the  time-sharing  method of
01900	operating computers and that the half-hearted measures that they have
02000	taken  in this direction have been bungled so that the present system
02100	is  not  cost  effective.    Moreover,  I  think  that  the  Stanford
02200	Computation  Center  will probably continue to sit on its hands until
02300	IBM offers a good time-sharing system and sends its salesment to push
02400	it  down their throat. Unfortunately, IBM will not get around to this
02500	for two or three more years, and by  that  time,  the  administrative
02600	itch  will have induced Stanford to commit itself to another obsolete
02700	system.  The following points are an elaboration of this position.
02800	
02900		1. A time-sharing computer system is one in  which  the  main
03000	interaction  of  the  user with his programs is via a console of some
03100	sort  rather  than  through  punched  cards.     A  general   purpose
03200	time-sharing system has the following features:
03300	
03400		a.   A  user  maintains  his files on the disk, edits, enters
03500	data, and runs his programs via his console.
03600	
03700		b. The user can conduct a dialog directly with his program.
03800	
03900		c. Interactive facilities are available for any  language  in
04000	the  system,  and programs of any size up to close to the size of the
04100	machine's memory can be interacted with.
04200	
04300		2.   The  present  Stanford  system  has   some   interactive
04400	capability, but this capability has the following limitations:
04500	
04600		a.   The  user  can  interact with the editor (WYLBUR) rather
04700	freely and with some special languages including LISP and  BASIC  but
04800	not  machine  language,  FORTRAN,  ALGOL  or PL/I.  Because memory is
04900	partitioned in a fixed way in the IBM operating  system  OS/360,  the
05000	storage available to interactive facilities is so small that only toy
05100	problems can be done.
05200	
05300		b. The interactive system is not cost-effective  because  the
05400	Computation   Center   has  bought  an  ineffective  and  excessively
05500	expensive terminal in the IBM 2741 (they should have used  the  Model
05600	33  teletype),  it  has bought a very expensive form of intra- campus
05700	data communication from the telephone company (they should rent  bare
05800	wires),  and  they  are  using  an  extremely  expensive input/output
05900	multiplexor (they should be using the PDP-9 computer they bought  for
06000	the purpose).
06100	
06200		You  might  imagine  that this situation merely reflects slow
06300	development of the technology of time-sharing.  This is not quite so,
06400	because   earlier   time-sharing  systems  did  not  have  the  above
06500	limitations, and the University of  Michigan,  using  equipment  very
06600	similar  to  Stanford's,  has  a proper time-sharing system that they
06700	developed themselves.
06800	
06900		3. The weaknesses of the main Stanford Computation Center are
07000	partially  responsible  for  the  fact  that computer facilities have
07100	proliferated all over the campus.    A  proper  central  time-sharing
07200	system  would  obviate  the need for many of the special centers.  On
07300	the other hand, if the Computation Center does not  provide  adequate
07400	time-sharing,  the  departments  and projects will be well advised to
07500	provide interactive facilities for themselves.  (In passing,  I  will
07600	mention  my  opinion that the ACME facility is a mistake and provides
07700	the Medical School with no facility that could not have been provided
07800	more  cheaply  centrally  and  has not prevented the proliferation of
07900	small computers within the Medical School).
08000	
08100		I  should  mention  that  I  have  been  involved  with   the
08200	development  of  time-sharing syystems from the beginning.  I suppose
08300	this gives me both a claim to expertness and the possibility of being
08400	biased.
08500	
08600		Professor   Miller's   memorandum   reflects   the  following
08700	situation:
08800	
08900		1. IBM has come out with a new computer system, the 370 which
09000	is  faster  than  the  360  models  presently  at  Stanford  and  the
09100	Computation Center will lose face if it doesn't get one.    According
09200	to  the  IBM salesmen, letters of intent have already been put in for
09300	several 370's, but this does't mean much without the cash.
09400	
09500		2. While IBM may offer proper time-sharing  systems  for  the
09600	370, it hasn't yet.
09700	
09800		3.   Stanford is short of money, and a major expenditure on a
09900	370 would require concentration of  resources.    In  particular,  it
10000	would  require  suppression  of  many  independent efforts to acquire
10100	computer facilities by projects and departments.  Miller's memorandum
10200	suggests such suppression.
10300	
10400		I  think  that  Stanford  should spend money on a new central
10500	computer facility only  in  connection  with  a  decisive  switch  to
10600	time-sharing.    If the Computation Center cannot present a detailed,
10700	convincing plan for doing this, Stanford should sit tight for another
10800	few  years.    This  would  mean  that departments and projects would
10900	continue  developing  their  own  interactive  computer   facilities.
11000	Because  of  the  particular importance of interactive computation in
11100	education, I would recommend that  the  Computer  Science  Department
11200	attempt to acquire an interactive facility for student use.
11300	
11400		One  new central facility which would be justified at present
11500	is a public mass storage facility accessible  from  any  computer  on
11600	campus.    This  has  been proposed for a number of years without any
11700	reaction, positive or negative, from the Computation  Center  and  is
11800	particularly  desirable  now that the IBM 3330 with high cost but low
11900	cost per bit has been announced.  (800,000,000 bytes rent  for  $7600
12000	per month.)
12100		I trust that the plans for expansion or  replacement  of  the
12200	central  campus facility will be made public and that opportunity for
12300	public discussion will be given before any new University  funds  are
12400	committed.    Another  blunder  here will have extremely serious long
12500	term consequences for Stanford.
12600	
12700		I have expressed myself rather vigorously in this  memorandum
12800	and  without  more than one conversation with Mr. Phillips.  However,
12900	he emits the same uninformative dusty  anwers  as  his  predecessors.
13000	Perhaps,  some computer science and other concerned graduate students
13100	might be  willing  to  hold  some  hearings  and  undertake  a  small
13200	independent study of the subject of Stanford's future in computation.
13300	
13400					John McCarthy
13500	
13600	
13700					Professor of Computer Science