Date: Sat, 3 Feb 2001 12:03:57 +0100 From: Rahul Siddharthan <rsidd@physics.iisc.ernet.in> To: Terry Lambert <tlambert@primenet.com> Cc: j mckitrick <jcm@FreeBSD-uk.eu.org>, freebsd-chat@FreeBSD.ORG Subject: Re: quote about open source Message-ID: <20010203120357.B94275@lpt.ens.fr> In-Reply-To: <200102022305.QAA16383@usr08.primenet.com>; from tlambert@primenet.com on Fri, Feb 02, 2001 at 11:05:58PM %2B0000 References: <20010202151744.O38235@lpt.ens.fr> <200102022305.QAA16383@usr08.primenet.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Terry Lambert said on Feb 2, 2001 at 23:05:58: > > By that reasoning, there can't be innovation in science either, > > because it's always been "open source" -- share your ideas, publish > > them, etc. The idea of patenting scientific discoveries is pretty > > recent, and even so, the equivalent of "closed source" (hide your > > methods, reveal only your results) just doesn't exist -- because > > people know such things cannot be taken seriously. > > You are aware that both Feynman and Dyson used Clifford Algebras > to do much of their ground breaking work, but did not share this > tool with their collegues, right? Most of the early QED (Quantum > Electro Dynamics) work they did appeared to "skip steps" for most > of their contemporaries, who lacked these tools. There was a lot > of "the proof is left to the student", which seemed like hand > waving. Even today, unless you take your QED classes from an > extremely enlightened professor, you are likely to not ever hear > about Clifford algebras, and even get a theoretical physics > degree, without laying your hands on this important tool. Well, I haven't read the early QED papers and don't know very much about the subject anyway (I'm not in high energy physics); I'll come back to the subject of Feynman below. But yes, what goes into the paper isn't necessarily how you yourself arrived at the results. However, the paper should make at least enough sense for the referee to believe the answers... It's pretty common even in "open source" for someone to work on something alone, without telling anyone, and then spring it on to the world when it's in a working state. The linux kernel is itself an example. My point was, you know a scientific claim works when you know all the steps of the theoretical reasoning or the experiment; you know a program works correctly (and no back doors, etc) when you have the source code in front of you. You are pointing out that there may be several ways to arrive at a scientific result, and scientists may keep some method to themselves. That is true, but they still publish *some* method to satisfy the rest of the world. I'm not sure there's an analogue for this in programming: if you want to verify that a program works as it should, you'll want the source to *that* program, not the source to some other program which gives you identical results. The correctness of computer programs can actually be a problem in physics too, given the reliance on computer simulations these days. Most people don't publish the source code to their simulation programs: the only way to verify the correctness of a simulation is for someone else to write their own code for the same problem, and compare the results. Coming back to Feynman -- his major contribution to physics was surely the idea of path integrals, which led to a diagrammatic way of doing perturbation theory, called "Feynman diagrams". Many of his results in QED were derived by Schwinger and others algebraically, but Feynman could do them much faster and much more intuitively. But the whole reason for accepting his methods was that he published them, and explained them, and they made sense to everyone. Compare a mathematician of the early 20th century: Srinivasa Ramanujan. He was not trained in the Western way, and had a habit of doing his calculations on a slate, rubbing them out, and only writing his results in a notebook; and people are still trying to prove those results today. His notebook was sufficiently interesting for people to make the effort to prove it all, but nothing he says can be trusted without proof. In a few cases he's been shown wrong, in fact. Now, the proofs that have been published may not be what he had in mind (and he may not have had a rigorous proof in mind at all), but any proof, if rigorous, will do. Most conventional mathematicians/scientists justify everything they say properly in their papers, but the arguments they actually write may not be the ones they originally used to arrive at their results. > I am aware of a number of physicists who are working on what > I consider absolutely brilliant ideas, and who have made very > significant progress in predicting phenomena, going down their > roads less travelled, but who are completely unwilling to share > their work, except in a very small circle of trust, until they > have pushed it much farther. One of these physicists literally > predicted the existance of the W particle _from theory_, and, > further, calculated its energy _on the nose_, back in the > early 1970's, and _still_ has only published bits and pieces. > > > As a more contemporary example, the Taniyama-Shimura conjecture > (that all elliptic curves have modular forms), which if proven, > proved Fermat's last theorem, was worked on by the mathematician > involved in total silence and isolation, until he believed he > had solved the problem. > > > Perhaps my favorite example is when Sir Edmund Halley went to > Newton, described his concept of a "comet", and asked Newton > to help him figure out what shape the orbit would take. Rather > than embarking on a long project over many years, Newton thought > about it for a few minutes, seemingly considering whether to > undertake a project that would cost him a large chunk of his > adult life, and stated "an ellipse". Newton, of course, did > not bother to tell Halley that he had earlier invented calculus. Newton did, of course, eventually publish his results. I'm not really familiar with the other stories. My point is, in science you nearly always have to publish compelling arguments to support any claim you make; and the more surprising the claim, the more thorough the arguments required. Your point about "trade secrets" which are not actually revealed in published work is correct, but as far as I am aware, in today's world at least, the published work should contain some alternative argument. Another reason to hide such "trade secrets" is that the people concerned may not be convinced about their correctness -- they are tricks that work, but their answers must be verified in other ways. Newton, for example, took many years to prove that two spheres attract each other in the same way as two point masses located at their centres; until he had a satisfactory proof, he did not want to publish it, but this was quite crucial to calculating planetary orbits and so on. Feynman had problems convincing older physicists of the correctness of the diagram approach (How can you draw a line representing an electron? What about the uncertainty principle?...) The works of people like L D Landau and P W Anderson (probably the people most responsible for whatever we know about "condensed matter" today) are generally brilliantly argued and very persuasive, but you get the feeling that they somehow "knew" all these things intuitively, and the arguments came only later. Maybe such people have "trade secrets" of their own. None of these situations seem to arise in programming. But the fundamental requirement, that in order to be satisfied about the correctness you have to see the inner working, is still there. And it is quite possible for someone to work alone, and "publish" the program only when sure of its correctness. - Rahul To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-chat" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20010203120357.B94275>