The Tech - Online EditionMIT's oldest and largest
newspaper & the first
newspaper published
on the web
Boston Weather: 74.0°F | Partly Cloudy

Computer experts meet to debate SDI feasibility

By Derek Chiou

Could reliable software ever be written for the Strategic Defense Initiative (SDI)? Five of the most prominent computer scientists in North America debated their positions concerning SDI Monday night in a packed Kresge Auditorium.

David L. Parnas of the University of Victoria in British Columbia and Joseph Weizenbaum of MIT, spoke against "Star Wars." Charles L. Seitz '65 of the California Institute of Technology (Caltech) and Danny Cohen of the University of Southern California (USC) spoke in favor of the system.

Michael L. Dertouzos PhD '64 of MIT moderated the discussion. The debate was sponsored by the MIT Laboratory for Computer Science (LCS) and Computer Professionals for Social Responsibility.

Parnas, a prominent software engineer and consultant for a number of US defense projects, was the first speaker. He was a part of the Strategic Defense Initiative Organization (SDIO) and a member of a computer research advising panel. He resigned from the panel, because he felt "SDI cannot meet its advertised goals," and because "SDIO is not a good mechanism for funding research."

Parnas claimed that reliable software for SDI could never be written. He said development of a successful "Star Wars" operating system is less likely than ten thousand monkeys randomly typing the Encyclopedia Britannica.

Software can be verified by mathematical analysis, case analysis, extensive testing or a combination of the three, according to Parnas. None of these methods apply to the SDI software, he said.

Parnas explained that "Star Wars" software is more difficult to write than other software because:

O+ it must locate enemy missiles that is not already known;

O+ it is impossible to ensure that it will not fail during an attack;

O+ it must meet "hard, real-time deadlines;"

O+ it cannot be modified during its actual operation nor can it be easily modified in peacetime because of its size;

O+ and it is susceptible to espionage.

The resulting unreliablity of the system renders it useless, Parnas said. He suggested it may even weaken our strategic defense by forcing potential opponents to speed up their own military build-up.

Weizenbaum, professor in the Department of Computer Science and Electrical Engineering and member of the LCS, supported Parnas' argument.

The specifications for the system are not and cannot be defined, according to Weizenbaum. The Soviet Union can always change its weapon systems; therefore, a program that will satisfy these unknown specifications cannot be written with any degree of accuracy. He shares Parnas's view that the program will not work when it is needed because it cannot be fully tested.

Weizenbaum also opposes the political aspects of the system. He said it is just a "technological fix of the grandest possible order" to a "political, social and cultural problem, in other words, a human problem." The arms race and threat of nuclear destruction is not a technical problem, Weizenbaum said.

Weizenbaum said the cost of SDI is too great. It is a step toward the militarization of space and society, he added, and it will escalate the arms race.

The first panelist to defend the feasibility of the "Star Wars" software was Seitz, professor of computer science at Caltech. Seitz's specialties include Very Large Scale Integration (VLSI) architecture and design.

The current objective of SDI is to conduct the vigorous research necessary to build a defense system, according to Seitz. He argued that such a system can be written using conventional software techniques coupled with radical hardware architecture. This will greatly aid in the testing, simulation and modification of SDI, he said.

Seitz said the EASTPORT panel, the computer advising wing of the SDIO, is examining a promising system whose components are able to function at least partially on their own if necessary.

The final panelist to speak was Cohen, director of the Systems Division at the USC/Information Sciences Institute. He has worked with the computer science departments of Harvard, Technion of Israel and Caltech.

Cohen addressed specific points of both Parnas's and Weizenbaum's speeches. He explained no one really knows the specifications of SDI and the research team must find them. Development of the software is not rendered impossible by difficulties in writing it, he said.

Everything that works now has bugs, Cohen said in response to Weizenbaum's statement about the present lack of specifications. Computer architecture will have a significant impact on making the entire system work, he concluded.

Speakers discuss feasibility

Parnas expressed disappointment with the pro-SDI speakers during the panel discussion. He claimed the SDIO was created to "develop" a system, not just to do research in the area.

He said independent testing of components would not be good enough because one part might interfere with another during actual operation. No matter how good the final system could work, it would still not free the nation of the fear of nuclear weapons. It would, therefore, not meet its objectives, he said.

Cohen countered that many systems will function if their individual components do not work, citing the telephone system as one example.

Seitz and Cohen reemphasized the research status of SDIO. Seitz said the military wanted SDIO's true opinions concerning the system's feasiblilty.

Parnas claimed a leader of SDIO had told him the Department of Defense would not accept advice to reject the entire project.

The problem of SDI's size was also discussed. The software necessary to manage "Star Wars" may be around three to five times larger than the biggest military program in existence, according to Department of Defense estimates.

Seitz said small, independent units are being investigated. Parnas replied smaller units are more independent but are weaker because of a reduction in communications.

Parnas said "Star Wars" would never be possible. The discrete functions would not be representable in a memory size acceptable for the system nor could they be analyzed efficiently, he argued. With SDI, "All of our eggs will be in one basket."

Seitz pointed out that although "Star Wars" would be untested and difficult to defend, all of the other existing anti-ballistic missle (ABM) systems are in the same position.

Evening closes with

individual statements

Judging from his extensive experience with computer programming, one can never trust software, Parnas said. The system would have no advantage and might even hurt the stategic position of the United States.

Weizenbaum called for the solutions to unanswered questions posed during the final minutes, such as submarine-lauched missiles which SDI would be ineffective against.

Seitz said everything is still in the research stage, and Cohen added SDI would be better than the ABM systems availible today.

Dertouzos, professor of computer science, director of the LCS and author of The Computer Age: A Twenty Year View, ended the debate with his closing comments.

Computer science is still a very experimental field and one never knows what kind of progress will occur in the future, he said. Though technology is advancing at a great rate, we are "the same ancient humans that built the Parthenon and the pyramids."