Is computer architecture too complex to be revolutionized?
Several of our reading’s have mentioned the absurdity of the fact that despite constant and consistent innovation in almost every other area of computation, the basic structure of computers has not changed at all. Peter G. Neumann, a computer architect old enough to have worked on the ill-fated multics system, suggest a complete redesign of computer architecture is needed; but, does computing need a revolution? This is not as cut and dry an issue as the article would make it out to be; this sort of redesign is even bigger than IBM’s “5 billion dollar gamble” with its System/360 line of products. The funds and technical knowledge required would be extraordinary with an unquantifiable and potentially smaller than anticipated payoff at the end of the process.
Who should (could, would) be working on this project?
Is computer architecture too important to leave to corporations? Is it a public good that needs to be cultivated with government funds to produce the innovations that will drive the next information revolution, or even strengthen the current one? Can we count on the government for not only large but consistent funding for a full restructuring of the basic computing architecture.
Even if we can somehow get the government to commit to a long term plan, how do we divvy up the results of this research? The natural solution seems like open source. Much like the publishing of the Architecture for the ENIAC prompted a series of copy-cats and improvements; the forthcoming research into new computing architecture needs to be publicly available and easily replicable. This will both speed design and allow for more complete error-checking ensuring the utmost security is maintained, the key driver of this project.
Who can (and will) do it?
It seems like there is only one viable path for this development, the defense R&D budget. Given the growing vulnerability military and governmental infrastructures face from cyber-terror and hacking, the justification for this spending seems like a sound investment. It should be much easier to convince the military brass to embrace the potential for a more secure computing platform than the political establishment of the future economic benefits of an incredibly large investment that won’t show returns for at least a decade. As the Manhattan Project, the IAS and DARPA provided the funds and the institutional support for nascent projects that ended up revolutionizing computing; so too can an investment in this sort of research provides the seeds for the next revolution (be it quantum computing, new utilizations of solid state drives, or parallel algorithms for use in personal computing).