COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\input buslet[1,ra]
C00007 ENDMK
C⊗;
\input buslet[1,ra]
\jmclet
\vskip 30pt
\address 
Editor
Stanford Daily
Stanford University
Stanford, CA 94305
\body
Dear Editor:

	The {\it Daily} report of the the anti-SDI quotes Winograd
as saying that people who understand computer systems agree that
the computer part of SDI can't work.  As a computer scientist, I
want to dissent from that.

	Discussing the specific problem of the SDI computer system
is not feasible now, because definite proposals haven't been made
about the physics.  For example, it isn't decided whether the system
would be ground-based or space-based or whether lasers or ``smart rocks''
or a combination would be used.  There also isn't a clear picture
about the extent to which the system will be centrally controlled and
to what extent control will be distributed.

	For this reason, Computer Professionals for Social Responsibility
and friends have had to resort to general arguments about the difficulty
of debugging large systems, and I will have to resort to general
arguments defending the possibility of debugging them.

	Examples have been cited about bugs in computer programs
controlling space vehicles.  However, let's ask to what extent
computer programming difficulties have contributed to delays in the
schedules, cost-overruns and disasters and what fraction of project
costs have involved computer programming.  My impression, and so
far as I know there has been no sytematic study of the matter, is
that computer programs have contributed a small fraction of the
difficulties, costs and disasters of these projects.  For this reason,
if computer programming is critical to the feasibility of SDI, there
is the possibility of putting much greater resources into it than
has seemed necessary in previous projects.

	Many large systems have worked the first time they have
been turned on including an enormous program for telephone switching.
Sometimes bugs have turned up years later.  The SDI system may function
perfectly in case of attack even though it has some residual bugs.  If
the programs are well tested, the remaining bugs are unlikely to be
excited in a single attack.

	Many will say that this won't do, that perfect safety is
required.  Alas, no strategy will get perfect safety, and SDI looks
worth exploring to see what improvements in our safety it will give
us.  I hope that Stanford scientists whose abilities are relevant
will help with SDI.

	Let me finish with an ad hominem remark.  The activists against
SDI are mainly the same people who have been activists against every
other defense program.  This leads one to the opinion that there is
nothing special about the computer problems.  It's just another thing
to argue about.

\closing
Sincerely,       

John McCarthy    
\annotations
\vskip 1in
%Enclosure
\vskip 1in
JMC/ra 
\endletter
\end