COMMENT ⊗   VALID 00005 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	sdi[f85,jmc]		Statement on SDI
C00004 00003	It is true that there have been system failures in the space program
C00005 00004	This is what I wrote before thinking that we might actually undertake
C00009 00005	NASA's Mannned Spacecraft Computers - James E. Tomayko, Annals of the
C00010 ENDMK
C⊗;
sdi[f85,jmc]		Statement on SDI

	This statement comments on various statements by computer
scientists and other computer professionals in opposition to the Strategic
Defense Initiative on the grounds that it requires computer programs that
are impossible to make sufficiently reliable.

	We fear that our colleagues are engaging in wishful thinking.
Namely, they are too ready to believe that computer science and
computer engineering support positions that they are inclined to
take for non-scientific reasons.

	Some of the statements imply that there are principles of
computer science that show that the required SDI programs are impossible.
Others are more cautious and merely state a ``professional opinion''.

	The arguments correctly point out difficulties but jump to the
conclusion that they are insuperable.  They argue that
It is true that there have been system failures in the space program
due to programming errors.  However, these have been enormously fewer
than failures due to hardware.  From this we can conclude that software
has not been so difficult to debug as to constitute the most expensive
or time-consuming part of the development of new systems.  This means
that if the software of SDI turns out to play a more important relative
role than software has in other space programs, there is a large
margin for the application of extra money and time to solve the
problem.
This is what I wrote before thinking that we might actually undertake
the relevant survey.  Of course, we need to check whether it has already
been done, but there is no trace of its having been done in any of the
pro or con arguments on the feasibility of the SDI software.

Whether the software problems of SDI would make it impracticable cannot
fully be estimated until a definite system is proposed.  However, some
relevant information can be obtained by answering the following questions
about previous large systems involving both software and hardware.

1. What fraction of system failures have been due to program bugs?

2. What fraction of delays in system availability have been due
to program bugs?

3. What fraction of cost over-runs have been due to underestimation
of the cost of software?

4. What fraction of total cost has been software?

	Answers to these questions can be obtained for NASA, DoD and
commercial systems and should help clear the air.  Lacking such answers,
let me mention some impressions based merely on having read the
newspapers.

1. While some examples of software failure have been publicized, they
are very small in number compared to hardware failure.  Moreover,
computer hardware failure is a small part of total hardware failure.
Certainly this is true of the Shuttle program.

2 and 3. The cited sources of delays and cost over-runs are almost
always hardware.

4. Software is relatively small fraction of the cost of most military
programs.

	In so far as past experience is relevant to SDI, this means that
there are likely to be resources for putting extra effort into software
if this is critical.  In previous programs, it would seem that the
competition for resources has prevented putting more money into software
than was required to keep software failures to the small fraction of
total failures that was actually experienced.

	These considerations bode well for SDI.
NASA's Mannned Spacecraft Computers - James E. Tomayko, Annals of the
History of Computing, vol. 7, Number 1, January 1985