COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "let.pub[let,jmc]" source
C00010 ENDMK
C⊗;
.require "let.pub[let,jmc]" source;
∂MLOT William F. Miller, Provost$John McCarthy, Director of LOTS$Opinions$∞

Dear Bill:

	Thank you for your kind memo.
Some of the events that prompted my declining the invitation to appear
before the study committee turned out to be unfortunate coincidences,
and anyway I decided to assert my point of view rather than just
withdraw, so I will come.  However, I should summarize my views
and tell you what made me unhappy and angry.

	I have long believed

	1. The best way to compute is interactively, and I invented
the concept of general time-sharing and initiated
the first work on it when I was at M.I.T.  I also
initiated the first display-based time-sharing system on the PDP-1
at Stanford and the first system based on having terminals in all
offices at the AI Lab.  I have always been disappointed that this
has never been considered to qualify me to
take part in Stanford's computer planning.

	2. The use of interactive and most other kinds of computing are
still substantially inhibited by cost.

	3. It is possible to organize a computing operation serving
a large user population and still keep personnel costs well below
the hardware costs.  I believe that LOTS with a paid staff of four and
about 3000 different users each quarter has demonstrated this.
While I underestimated the amount of computing power required to
switch student use to an interactive mode, we are getting by.  Moreover,
the main weakness of LOTS is not enough computer rather than an
insufficient staff.

	Let me assure you that I have not been a figurehead at LOTS
either.  I have initiated most of the major activities, and I have
made the major policy decisions.  Ralph has done an excellent job,
but if he leaves for a job with greater responsibility, we
will be able to replace him.  (We may need another person if
we can't get someone as versatile).

	4. Stanford could meet its general computing needs with
many fewer people than operate SCIP.  Moreover, it could have
switched its operations to interactive computing much faster than
it did.

	5. While there is nothing that intrinsically prevents a
centralized organization from being efficient, centralizing removes
some of the inhibitions against organizational inflation.  When
engineers, physicists and computer scientists provide for their
own computing needs, they are unlikely to overstaff computing
at the cost of hiring engineers, physicists and computer
scientists.  Given the bureaucratic ethos of the 1960s and 1970s,
it would have taken very strong management for SCIP to have
escaped organizational inflation.

	While decentralization would probably make scientific
and educational computing more efficient, different methods
may be required to reduce the cost of administrative computing.
Maybe combining decentralization, the use of outside services
and across-the-board budget cuts would work.

	These general views are not based on experience with large organizations
or administrative computing and may be discounted in deciding what has to
be done to serve a large community and how to do it with as few people as
possible.  On the other hand, the general views were the basis for setting
up LOTS and running it.

	I was disappointed that you did not consider LOTS and the
AI Lab sufficiently successful to warrant
including me in your study, but that is your prerogative.  What made me angry
was to receive an invitation over your signature to the hearing that did
not include me as Director of LOTS or even a phone call telling me why not.
When the external Advisory Committee invitation also did
not include me, I mistakenly concluded that there was a positive desire to
avoid my point of view.

	I have since learned that the list for the Advisory Committee was
made by Feigenbaum's new secretary, and the Committee made amends by
inviting me to dinner.  This encouraged me to call Jon Sandelin and tell
him that I would speak for LOTS unless specifically
asked not to, and he apologized for what he said was his own mistake.

	Ralph has done an excellent job as Manager of LOTS, and when
occasions for publicity have arisen, I have been content to stay at
the AI Lab and let him be interviewed.  Nevertheless, I have fulfilled my
responsibilities as Director, and I have ideas and proposals for the
future of LOTS that I would like to implement
once we have solved our usage crunch.

	Even with the clarification that has occurred, I still am not
confident that my views on the future of computing at Stanford in general
or even at LOTS are sufficiently wanted so that it won't
waste your time, the committee's time and mine to express them.

.reg