COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	sdi.2[e85,jmc]		Statement on SDI
C00006 ENDMK
C⊗;
sdi.2[e85,jmc]		Statement on SDI

	According to Science (1985 July 26, News and Comment), ``The
resignation of a highly regarded consultant to the Pentagon's
`Star Wars' program has brought to light a controversy within
the computer science community over the program's feasibility.
Charging that {\it the goal of the program is unattainable because of
inherent limitations in software reliability}, David Parnas \ldots''.
{\it Science} goes on to quote Anthony Ralston as saying ``Quite
apart form any other technical, political or economic objections
which might be raised about the Star Wars system, its software
problems doom it to failure, \ldots.  In no forseeable future
\ldots is there any valid prospect of writing 10 million or
100 million or anything approaching this number of correct lines
of code''.

	There may be a controversy within the computer science
community, but it's not about computer science.

	In our opinion there is no basis in computer science for
asserting ``inherent limitations on software reliability''.  There
is no principle of computer science that says 10 million or 100
million or any other specific number of lines of correct code
cannot be written.

	There is relevant experience --- some good some bad.  Some large
software projects have failed, and some have succeeded.  An example of
success is the AT&T electronic switching program containing close to 50
million lines of code.  Success depends on the usual factors of having a
clear plan, adequate management, sufficient resources and basic technical
feasibity.  When the proposed SDI system is more fully defined, its
computer and software requirements can be determined and the cost of
realizing them determined.  In the area of making large systems reliable,
especially systems which must function correctly under conditions where
ordinary debugging is too expensive or dangerous, there are methods, both
mathematical and using simulations, whose possibilities have not been
fully explored.

	We fear that our colleagues have exaggerated
the relevance of their technical knowledge and experience in order to
support the anti-SDI cause.