University of Newcastle, NSW
@10 June 1986
We kidded ourselves for a while that Star Wars had gone away. We pretended that flower power was winning. But in our heart of hearts, in our 3 a.m. nightmares, we knew that no toy of destruction, once conceived of, has ever been left to rest. Like Mordor’s ninth ring of power*, hidden forever deep in a dark river beneath a mountain, some Gollum was sure to chance upon it, and once set free it would again corrupt all who carelessly picked it up; (*J.R. Tolkien, The Lord of the Rings). Below is a computer engineer’s report on the technical viability of Star Wars I, circa 1986. Judge for yourself how relevant it is to Star Wars II, millennium edition…
The notes to follow are my best attempt in 1986 as a non-specialist to outline a lecture given by David Lorge Parnas , a Canadian computer engineer and professor who spent years working on American defense projects. Parnas seemed uniquely qualified to report on the Strategic Defence Initiative (S.D.I.). At the time I found him highly persuasive, and nothing since, to my knowledge, has invalidated his technical arguments. For that reason, it seems worthwhile to re-present them here, a generation later. Any errors of interpretation in this document are of course my own, and readers with a more substantial interest are advised to read David Parnas’s own collected essays on the topic, which are now available online .
In his lecture, David Parnas was not mounting a political argument against the Strategic Defence Initiative, but illustrating why, technically, it was not feasible. Of course, his technical discussion had, and has, intense ethical and political consequences.
We know now that the military superpower of the USSR, so fearful in his generation, did collapse politically. The reasons for that collapse are complex, but military overspending certainly played a large part. Whatever the technical infeasibility of the S.D.I., the misallocation of resources which it excited from technically illiterate politicians extracted its own consequences. It just happened that the United States of America at that time had deeper pockets than the Union of Soviet Socialist Republics. We see now that the hubris emanating from American “triumph” in the arms race has since also claimed America as a victim, an historical pattern familiar from almost every empire in history. Oh, and those Russian missiles which the S.D.I. was supposed to render obsolete are still there, and an invulnerable “space shield” is still a wet dream for the big spenders in Washington.
As a linguist I have a special reason to remember this lecture by David Parnas. It is an odd fact of intellectual insight that sudden understanding can come from unexpected directions. A key element of Parnas’s discussion is that computer engineering deals with discrete, randomly discontinuous phenomena whose states are too numerous and variable to model mathematically, and whose outcomes can therefore not be precisely predicted. Conventional engineering deals with analogue phenomena whose infinite states form a continuum which can be reliably modeled and tested. For example, a structural engineer can be fairly confident that a model bridge which he has tested in a wind tunnel can represent a similar bridge scaled to full size. A computer program cannot be scaled in this way.
The insight which came to me after David Parnas’s lecture was that natural language, like a computer program, is also a kind of randomly discontinuous phenomenon which cannot be reliably scaled. This directly contradicted the logic-based Chomskyan generative model of grammar which I was using at the time as the base for a doctoral dissertation on grammatical agency. I eventually abandoned that dissertation after deciding that in principle natural language patterns could not be accounted for by a logical generative approach. My doctoral withdrawal then was not a decision which found much sympathy or understanding, and any professional linguists reading this will know that there is still a happy tribe following (somewhat evolved) forms of generative grammars. They will be there until hell freezes over. Careers and reputations are involved. The same forces of natural reaction will of course be at work in Parnas’s field of computer engineering, and without doubt in the universe of weapons development.
The Strategic Defense Initiative (S.D.I. or “Starwars” Project) – Why It Won’t Work
Notes on a lecture by Professor David Lorge Parnas
.. delivered under the auspices of the Australian Society of Engineers in Newcastle Town Hall, 8pm, 10th June 1986
David Parnas is a computer engineer who spends most of his time working on high technology United States Defense Department projects (e.g. the S7 aircraft).
He was invited to join a technical panel on the S.D.I. project and paid, like the other consultants, $1,000 per day [ed. not bad money in 1986]
His own inquiries led him to the conclusion that the S.D.I. would not and could not work. His fundamental objections related not to the physics of the exercise (which were also doubtful) but to the insurmountable human limitations on error-free computer coding.
On resigning he submitted eight technical papers detailing his objections. These have been widely circulated.
He was replaced on the panel by a retired general. No computer scientist would accept the position.
2. The Special Nature of Computer Engineering
Conventional engineering projects may be tested by:
a) mathematical modeling
b) multiple case testing
Computer programs cannot be tested by mathematical modeling because it is not possible to write equations for randomly discontinuous phenomena.
This is why computer code is always the last element of any large project to come right – often years after the hardware has been perfected. Computer code can only be made reliable by continuous adjustment in actual use.
There is evidence that complex code can sometimes become unstable through debugging – the elimination of one error leads to the generation of many more.
3. The Particular Problem of Coding for the S.D.I.
The S.D.I. may involve 10 million lines of code.
Multiple case testing of the S.D.I. is obviously not possible. The first test will be the last (i.e. warfare).
Because computer codes are randomly discontinuous phenomena, their failure characteristics cannot be derived statistically. That is, a failure/error in one part of the code cannot be generalized to other parts of the code.
A program may run with a thousand errors, but a single crucial error can lead to total failure. A few years ago, a Venus probe was destroyed because a semi-colon was replaced by a comma in a program.
Software controlled operations therefore have an element of inherent unpredictability.
4. The Operational Reality of the S.D.I.
The S.D.I. program will involve hundreds of satellites. Each satellite must contain one or more computers. Under battle conditions, the computers must run in real time. If the system is networked, one weak satellite link threatens the integrity of the whole operation. For example, if a signal is delayed by jamming at one point, this will distort computation in neighbouring satellites which are deciding when, where, how, and if to fire.
The “Byzantine Model” and other theoretical models show that if a networked real time system is split even at one point, the root nodes will never be able to synchronize.
If each satellite were autonomous, it would have to contain the whole S.D.I. program – there is no way to scale for a reduced version of the code. Practically however, in an operational theatre satellites cannot be autonomous since the operation of their platform weapons “blinds” their electronic reception. That is, “hits” must be verified and computed by neighbouring satellites.
Laser weapons do not “explode” anything. All they can do is to disable the guidance and other electronic systems of a missile (which will continue on course).
Since electronic weapons systems have no measurable signature, dead or alive, it is difficult or impossible to know whether a hit has been scored. This means that the satellite weapons platform doesn’t know whether to re-attack, or whether to direct action to other targets. Further, decoys, cheap and in huge numbers, may be deployed and be effectively indistinguishable from real targets.
5. The Security of the S.D.I.
There is no such thing as a secure computer code. Hardware can be locked up. Software can be encrypted, not merely for protection but for theft, in an infinite variety of ways. For example, a program can be encrypted in a “picture”, transmitted in multifarious forms by telephone, and so on.
Weapons software in the hands of an enemy can be subverted or circumvented. For example, there are cases where states in conflict have purchased the same weapons system. In a recorded instance, one protagonist used its knowledge of a shared radar program to reprogram war planes to fly in patterns which would be “invisible” to the enemy. So much for the public proposal by President Reagan to “share” the S.D.I. technology with the Soviet Union.
The Star Wars code can never be secure. Further, the large number of personnel involved means that one corrupted individual can easily write a devastating but invisible “logic bomb” or “worm” into the code.
In the whole history of computing NO system has ever worked on the first try. Even the lunar landing had to be done on manual override because the computer was going to dump the module in a crater. The S.D.I. HAS TO work in its entirety on the first try.
The Soviets have no space based S.D.I. program [ed. circa 1986]. They have a primitive land based program of a kind abandoned by the United States. From a computer programming point of view, land based systems are infinitely less complicated than computing for moving weapons systems in space.
The S.D.I. will lead to arms acceleration because military strategists ALWAYS work on a worst-case scenario. The U.S. military must assume that that the S.D.I. won’t work, and preserve their existing weapons systems. Soviet strategists must assume that the S.D.I. may work at least partially, and multiply their missile throw power. This generates a new arms race.
Defenders of the S.D.I. are now saying that it will not be used to protect civilian targets, but only missile silos. However, hardened missile silos are now invulnerable to a strike within 500 feet. It is vastly cheaper and more reliable to defend that 500 foot radius by a ground based system than from space.
6. The Commercial and Political Dimensions of the S.D.I. Development
Contracts for the S.D.I. are being deliberately distributed throughout the states of the American Union, and amongst international allies. This will make it politically difficult for a succeeding President to close the project down.
What are the economic and technical advantages of participating in the S.D.I. projects?
a) Technical Knowhow: With all Defense Department subcontracting, knowhow passes up the hierarchy from the subcontractor to the central bureaucracy, but an absolute minimum amount of information to do the job (and often less than that) is passed down the tree.
b) Product Spinoffs: An inseparable component of engineering feasibility is cost. The cost structure of defense related products is so remote from the marketplace that in fact very little technological transfer takes place. All large Defense Department contractors have separate divisions for military and commercial activity which, for commercial reasons, remain largely unrelated. If a job costs an individual $1, it will cost a small company $2, a large company $20, and the Defense Department $200. In any case, most defense related products, in their very nature, are unusable in the civilian sphere.
7. The Defense Perspective of the Soviet Union
Why do the Soviets appear to be upset about the S.D.I. ? Parnas thinks that they may realize that with its massive diversion of skills and funds, the S.D.I. could set back the defense preparations of the United States by a generation. He thinks the Soviet Union may be working on the “Brere Rabbit psychology” of “yell NO loudly enough until the other guy goes out and commits foolishness from pure spite”. [ed. Well we know now that they were sold on a different fable, and followed Alice down the rabbit hole].
8. Miscellaneous Comments by David Parnas
a) The peace movement is a waste of time, says Parnas. If 60,000 people march in a city they do nothing but antagonize some shopkeepers. TV audiences watch with half their brains – they are not activated. If 60,000 people each go and persuade 5 doubtful neigbours, THEN there has been some gain.
b) Effective opposition to the S.D.I. will come through word of mouth only.
c) President Reagan is only told what he wants to hear by his minders. He receives no reliable scientific advice at all.
d) Freud (writing to Einstein): Law and Force are equivalent phenomena. Law comes about because individually weak units band together to control a stronger element.
e) Parnas: This fact holds for states as well as individuals. The superpowers will never act unilaterally to curb their own power. Relatively weak states such as Australia and Canada may be able to act in unison to bring about an international rule of law which constrains the superpowers.
f) There are characteristics of inefficiency and corruption which are common to all large organizations, irrespective of ideology or nationality.
g) Most of the scientists participating in the Strategic Defense Initiative are knowingly perpetrating a fraud.
h) Social courage is rare. For example, the engineers who knew that the space shuttle, Challenger, was at risk didn’t go public until after the event.
i) Military programming technology is generally inferior to that in the private sector. For example, the Defense Department Programming Language Project, ADA, is a monster that has never worked, but is not allowed to die. Contractors are required to work with it, wasting endless time and money.
j) The NASA ground control computer programs are a rat’s nest whose interdependent properties nobody understands. They break down constantly. (This information is based on Parnas’s own consulting experience). To get the S.D.I. satellites into space, the military would have to commandeer the whole civilian space program for a long time to come.
 Wikipedia (accessed January 2013): David Lorge Parnas @ http://en.wikipedia.org/wiki/David_Parnas
 Parnas, David (1985) Software Aspects of Strategic Defence Systems. American Scientist, Journal of Sigma Xi, Vol. 73, No. 5, pp. 432–449. Reprinted online in Communications of the ACM, December 1985, Vol. 28, No. 12 @ http://www.cs.helsinki.fi/u/przybils/courses/CBD06/papers/p1326-parnas.pdf
Professional bio: Thor May’s PhD dissertation, Language Tangle, dealt with language teaching productivity. Thor has been teaching English to non-native speakers, training teachers and lecturing linguistics, since 1976. This work has taken him to seven countries in Oceania and East Asia, mostly with tertiary students, but with a couple of detours to teach secondary students and young children. He has trained teachers in Australia, Fiji and South Korea. In an earlier life, prior to becoming a teacher, he had a decade of drifting through unskilled jobs in Australia, New Zealand and finally England (after backpacking across Asia in 1972).
contact: http://thormay.net thormay AT yahoo.com
All opinions expressed in this paper are entirely those of the author, who has no aim to influence, proselytize or persuade others to a point of view. He is pleased if his writing generates reflection in readers, either for or against the sentiment of the argument.
“ A Collision of Technology and Politics – Star Wars Revisited ”
© copyrighted to Thor May; all rights reserved 1986