COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\\M0BASL30\M1BASI30\M2NGR30\M3CLAR30\M4NGR40\.
C00014 ENDMK
C⊗;
\\M0BASL30;\M1BASI30;\M2NGR30;\M3CLAR30;\M4NGR40;\.



\C\F4COMMENTS ON TECHNICAL REVIEW OF XEROX PARC\F0

\CJohn McCarthy, November 1974

\C\F3SYSTEMS SCIENCES LABORATORY\F0


POLOS

\J	The time is ripe to develop office communication systems as
commercial products, and Xerox seems to be well in advance of other
companies in realizing it.  We have had such facilities at the Stanford
University Artificial Intelligence Laboratory for several years, and
on the basis of our experience, the essential facilities are the
following:

	1. Each secretary and each other person who will send and
receive messages or who will originate, modify, comment on, or read
documents produced in the system should have his very own terminal.
In laboratories where the essential facilities exist but where terminals
have to be shared the actual use of the facilities is quite limited.
(These include many of the TENEX sites on the ARPA net.)  The utility of
the system depends on how many of the people with whom one must interact
are similarly equipped.  Therefore, many people who will use the system
seldom must have terminals for the benefit of those who communicate with
them.  Therefore, keeping the cost of the terminal down is essential.
Since the terminal must also be tolerable in shared offices, it pretty
well has to be display based.

	2. There must be convenient \F1editor\F0 programs for preparing
and editing documents, and there must be a file system capacious enough
to retain all documents prepared in the system.

	3. Printing facilities good enough for letters and reports are
very important.  Having the XGP printer has made a big difference to
us.

	4. The last requirement is a facility for sending and receiving
messages and keeping track of them.

	5. On such a basis, a superstructure of information and document
retrieval programs will be erected provided inflexibility is not built
into the system.

	All these facilities must be convenient to learn and easy to use,
and it is not yet clear what is good and bad in this respect.
Secretaries may present less of a problem than casual users.  Some systems
are easy to use after a three months hiatus, while others present 
what I call the \F1instrument flying syndrome\F0 - if you haven't done
it for six hours in the last ninety days, you are no longer qualified.

	The POLOS project is aimed at an advanced office communication
system, and at a facility for doing research in office communication with
a rather small experimental system to be installed in the administrative
offices of PARC.

	The plans for the project seem entirely feasible and there is no
serious doubt in my mind that the project will be a technical success in
that the secretaries and administrators who use it will find it superior
to their present way of doing things.  Indeed, much more limited systems
already meet with good acceptance.

	However, \F2the system seems to me to be too elaborate in its equipment
for an initial office system, and I have some doubts that the advanced
features are those will be wanted by the users of simple systems\F0.
Specifically, I think the voice switching and storage facilities, the mouse,
the auxiliary keyboard, and the high resolution of the display contribute
more to the expense than they are likely to be worth in the forseeable
future.  (Let me admit here that I have long been one of the main proponents
of pure keyboard editing in opposition to those who consider pointers
essential).  Expense is very important, because the utility of an office
system depends on how widespread it is.  I find it frustrating that I
can communicate freely with people all over the ARPA net, but can't send
messages as easily to many people on the Stanford campus.
On the internal level, I think that the decision to use multi-processors
is more technically interesting than economically justified.

	To focus my concern more sharply, it may turn out that Xerox will
spend more money on R and D in office communications systems than any
other company but will be beaten to the big market by someone who aims
more accurately at the essentials.  I was assured that this isn't possible
because of developments elsewhere in Xerox, but I am not quite convinced.

	The advanced office system seems rather too like the POLOS to
me except for the work on interactive composition.  This is a nice project
and I expect it to succeed, although I would like to see the non-interactive
composition system that Xerox already has pushed faster towards actual
use.  We and Carnegie-Mellon and M.I.T. already use it.

	The examples given in the office graphics presentation didn't
convince me that this facility is going to be worth a lot except as an
add-on.


INFORMATION PROCESSING PSYCHOLOGY

	On the whole, this work seemed to be well-conceived, and I would
like to see its methods applied to comparing alternate methods of editing
such as comparing keyboards and pointers.\.


\C\F3THE COMPUTER SCIENCE LABORATORY\F0

\J	The Mesa programming system could not in the time available have
been presented in sufficient technical detail for me to evaluate it,
especially as I am not sufficiently familiar with its competitors.  What
I did understand looked good.

	The research on program analysis by Wegbreit, Deutsch and Moore
is quite well conceived and should make a substantial contribution to
computer science.  However, I think they are over-optimistic about how
soon the work will have practical applications.  In our own lab, we are
following an approach that relies more on machine assisted human proofs
of correctness of programs than on unaided machine proofs.  I think this
is a necessary first step towards practical systems, but it is also worthwhile
to explore the limits of the present techniques of automatic verification.

	I found the report on the \F1Production of Software\F0 quite
unconvincing, especially the comparison with the way IBM is alleged to
do it.  Many more and better controlled experiments are necessary
to reach valid conclusions, and the investigator seems to have jumped
to the conclusion that he has already proved his way best.\.


\C\F3GENERAL REMARKS\F0

\J	The Computer Science Laboratory is very strong scientifically.
Xerox has succeeded in getting some of the very best young people in
computer science.  They are still working quite independently of one
another, many on just what they were doing before they came to Xerox,
and their efforts have yet to be focussed on a small number of problems.
It seems to me that this will come with time, but it would come sooner
if there were a stronger scientific leadership.  However, all the problems
being worked on are worthwhile and will yield results of scientific value.

	The Systems Science Laboratory is working on concrete projects all
of which seem likely to achieve technical success in that the systems
will meet their specifications.  However, the scientific rationale of some
of the projects is weak, and I have some doubts about whether some of
them will result in products that Xerox will want to market or even
materially contribute to the development of such products.
Again the situation can be improved by better scientific leadership.\.