COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	∂16-Nov-86  1808	JMC  	re:      SDI debate
C00015 ENDMK
C⊗;
∂16-Nov-86  1808	JMC  	re:      SDI debate
To:   6058598%PUCC.BITNET@WISCVM.WISC.EDU  
[In reply to message sent Wed, 12 Nov 86 19:47:45 EST.]
See f86.in[let,jmc]/316p

Sorry to be slow in answering your questions.

First let me tell you the story of my involvement in the controversy.  In
1970 or thereabouts I was a good liberal and signed a petition promoted by
Dan McCracken against ABM.  I later asked another faculty member somewhat
to the left of me whether he had signed it, and he answered no, and said
that he didn't think George Forsythe knew what software was or not
feasible and neither did he.  I said "My goodness, is that what I signed?"
and promptly wrote McCracken withdrawing my signature.  In 1983 or 1984 I
had some dealings with Robert Jastrow on another matter, and he later
asked me to sign a statement in favor of SDI.  I declined but said that if
anyone claimed that some principles of computer science proved that SDI
was impossible I would sign a counter statement.  I fully expected to be
called on this, and I was.

I have not at any time been involved in research aimed at SDI nor in
any of the committees dealing with the relevant computer problems.  I
have no objections but haven't been asked and haven't volunteered.

Now to expound my position.  First the non-computer part.

1. I'm in favor of the objectives of SDI.  I believe that some defense
is better than none, so my support doesn't depend on any precise estimate
of what amount of protection can be obtained.

2. I should mention also that I have had dealings with Edward Teller in
another (non defense) connection, and became an admirer.  I have some
slight bias in favor of proposals he favors.

3. I don't believe the arguments that SDI is destabilizing.  I don't
agree that game-theoretic models of the military situation of the U.S.
and the Soviet Union cover enough of the facts to provide a basis for
analysis.  No-one is likely to attack on the basis of a now-or-never
analysis of the military situation, because the military facts will
always be too unclear.

4. I regard the largest probability of nuclear war as residing in
the underdeveloped war where the hostilities are greater and the
rationality is less than in the U.S. or even the Soviet Union.

5. I don't regard the U.S. and the Soviet Union symmetrically.
If the Soviet Union were a democracy, albeit a socialist democracy,
I wouldn't fear them.  Democracies don't go to war with each other.

6. The largest danger of war with the Soviet Union arises from the
possibility that their opaque and conspiratorial politics will bring a
military adventurer to the top.  Another (less likely) possibility is that
they will mistakenly come to believe from U.S. weakness and internal
politics that the U.S. can be bullied in some matter.  Historically
oligarchies generate political surprises.  I can't evaluate the
probability, but better safe than sorry, and to my mind, adequate U.S.
military strength is safer than the reverse.

7. Finally, I regard the SDI "technical" debate as a continuation of
previous defense "technical" debates.  The same players are on the
same sides, and this seems unlikely to be a co-incidence arising merely
from their accidentally held scientific opinions.  Some people argue
for various defense measures consistently, and others argue consistently
against them.  There is a slight asymmetry arising from limited resources.
The advocates of certain defense measures usually have to argue against
certain others.

8. Many of the computer opponents of SDI seem to be substantially
motivated by wishful thinking.  Bluntly they hope SDI will be found
to be infeasible.  Why they should hope that is unclear.  Mainly I
think it's a continuation of politics by other means.  Thus many of
the people opposing SDI allegedly for computer professional reasons
were members of Computer Professionals for Social Responsibility who
opposed the Pershing missiles giving reasons having little to do
with computing per se.

I think this is mainly a continuation of American defense politics
in which technicla arguments were made on previous defense issues,
but the friendships and alliances involved clearly had a political
basis.  People on the left often hate Reagan but regard Gorbachev
as a kind of force of nature - no more an appropriate target of
animosity than Mt. St. Helens.

9. I regard efforts at inducing a mood of desperation about the likelihood
of nuclear war as tending to increase the probability of war rather than
reducing it.  Namely, desperation induces oscillations in politics and of
policy.  Carter's reference to "inordinate fears of communism" and
embracing Brezhnev led the public to replace him by a President who
referred to "the focus of evil in the world".

10. In writing a thesis about the matter you will be kidding yourself
if you attempt to interpret the arguments by either side as totally
motivated by technical considerations.

Now to computer issues.

1. In general, I would like to reserve my position on the feasibility of
the computer part of SDI until a specific system is proposed.  I can
imagine concluding that a specific proposal is defective or that the
arrangements for implementing it are defective.

2. I don't believe there is any principle of computer science from which
it can be concluded that a complicated system can't be built with a
high probability of working the first time.  I have read the arguments
by David Parnas and by David Redell and Greg Nelson, and they seem to
have a purely qualitative character.  They identify some difficulties
with SDI systems, and say they cannot be overcome.  Opponents agree
that these are difficulties but expect to overcome them.

3. A substantial reason for hoping that the computer problems of SDI can
be overcome is the small part that computer hardware or software problems
have played in the delays, cost-overruns and disasters of previous computer-
dependent systems.  This is certainly true of the Shuttle program, even
though one or two launches were delayed a few days by computer problems.
If computing unreliability turns out to be a substantial part of
the difficulties with SDI, enormously more effort can be put into it
than has been required by previous systems.  [It is possible that
some other systems have had most of their problems stemming from
computing difficulties, but I haven't heard it alleged.  A possible
example is the Aegis missile ship system].

4. I see no reason why previously used methods won't be sufficiently
successful in ensuring adequate reliability of SDI computing.  Examination
of each bit of code by many people, identifying critical parts and
extensive debugging using simulation can all be used more intensively
than before.

5. The SDI system would probably function with some remaining bugs.  It
will be used once, and the particular attack that occurs may not activate
the more obscure bugs.  At least the bugs that survive a long time in
other systems are those which don't prevent early runs from being successful.

6. SDI is a military system, and throughout any military conflict and especially
in the beginning, there are all kinds of failures both in offense and
defense.

7. I would like to see program verification technology tried out on
suitable parts of SDI, because I think it would increase its reliability.
I don't think it's absolutely necessary, however.

8. I have heard that some people have advocated use of AI in SDI, but
I doubt they have proposed anything concrete.  I would want to see
any such concrete proposal examined very carefully.  Certainly I don't
see any present requirement for it, but it might help deal with certain
kinds of unexpected situations.

Well now, sorry to have been so lengthy.