pratt@cs.stanford.edu (01/18/91)
The overseas announcements have been useful to me. Though I've never gone to an overseas concurrency conference (well, hardly ever, there was the GMD-sponsored workshop on "Combining Compositionality and Concurrency" in Koenigswinter in Feb. 1988), I keep putting them on my calendar as I see them advertised here in the hope that I will find the time one of these days. From: Steven Ericsson Zenith <zenith@ensmp.fr> it is likely to further isolate the european parallel processing community which still hasn't woken up to the fact that our friends in the USA caught us with our pants down and now possess technology and resource in high performance computing far in advance of anything Europe currently possesses or is likely to possess in the next 10 years. Sequential ironmongers have traditionally made more money than the software vendors, though sequential software does seem to be getting its act together. However any sequential ironmongers who don't have ten years of in-house software expertise, or at least a very healthy third party vendor program (but not necessarily orchestrated by them, witness PC's), are in a very weak position today. Concurrency magnifies this effect ten-fold. Whoever gets the upper hand in concurrent software will get the upper hand in concurrent computing. While I don't think Steven is intentionally pulling the wool over our eyes, the Europeans are so far ahead of the US in concurrent software that in ten years time they will be the ones dictating terms to the hardware companies in the US. The situation with concurrent computing in Europe will be like that for automobiles and electronics in Japan today. Just as the Japanese added quality to manufacturing, to their enormous financial benefit, so will Europe in ten years time have added software to concurrent hardware. Reverse engineering the basics is no trick, the trick is to add value. Adding quality to automobiles and electronics was evidently hard, or the US would have done it as soon as they felt the heat. Reverse engineering the concurrent hardware the US is presently developing will similarly be no trick. I can assure you from 25 years of thinking about both sequential and concurrent computation that adding concurrent software isn't any easier than adding quality. In fact it is more than a trick, it is the ultimate martial art, one that can be mastered only with time, patience, and insight. The US learnt its lesson from the Japanese the hard way. If it learns the analogous lesson from Europe the hard way I'd say sell the US short now. For myself I'm hanging in here since I like California (reminds of Australia except the SF beaches are colder), but we do have some serious studying of the situation to do. Vaughan Pratt