On Mon, 30 Oct 2006 08:00:02 -0800, Bill Hart <[EMAIL PROTECTED]> wrote:
> > Google groups seems to have just lost a substantial reply I made. This > has happened more than once and seems to be related to the "reply > without quote" option. > > I don't want to type all that again, but here are the salient points: > > 1) MAPLE are pretty careful about their code. MAPLE authors also tend > to be so. I once saw a very neat trick that had been implemented in > MAPLE for the quadratic sieve. It seemed obvious, once you knew the > trick. It had even been written up in a paper. Ordinarily I could then > just use that trick and mention which paper I got it from. But in this > case it had been pointed out to me by a friend of the author. > Apparently the author didn't necessarily want his idea GPL'd. Better > suggestions are: I think I read the long reply you made just now, so I don't think it really got dropped. "I had a similar problem with the quadratic sieve I was developing. Someone developed an extremely nice trick for speeding up the quadratic sieve, implemented it in MAPLE (I'm not sure if it is actually part of the core functionality of MAPLE, since it seems to factor extremely slowly), then wrote a paper on the subject. I'd really like to pinch that algorithmic trick, and in fact, had I discovered the paper on the web myself, " > --Find a similar trick implemented in GPL'd software. > --Find a trick implemented in source that has no restrictions. > --Come up with your own trick which achieves a similar result. > > I have some questions: > > 2) Can one make use of code which has no restrictions, but which is > clearly copyrighted and maintained by a particular identifiable > individual, and GPL the resulting package? I note for example that some > of Jason P's msieve code (which has no restrictions apart from > attribution and an encouragement to notify him) has ended up in GGNFS > with attribution (though I can't rule out that he contributed it > himself). For example, would all of SAGE be GPL'd if it included work > from someone which wasn't GPL'd but had no substantial license > restrictions? Basically, I think the answer is "yes". Google for "GPL compatible". > [Unfortunately we can't just make use of msieve in SAGE, since it is > (currently) only fast on Athlons and that by virtue of its home brew MP > library which has a heavy assembly language component, I believe. Jason > P is an awfully nice guy, and the code is extremely fast, so otherwise > this might have been possible.] > > 3) How does one attribute the work of another GPL'd author when one > produces a derived version which isn't just a straight modification of > the original code. For example in one small section of my quadratic > sieve (the F2Matrix code) a small portion of it is based on, but not > the same as the F2Matrix code in Pari. My version is 20 times faster on > some architectures and the underlying algorithm is quite different but > some (not all) functions are still recognizable, though almost no two > lines are the same. > > I've tried contacting Gerhard Niklasch a number of times recently > without success. He appears to have written the code I speak of, or at > least is the current maintainer. > > The specific question I have is, what copyright do I put on the code? > Should I copyright the code I have put together to me and add a note > that I based part of it on their work, or should I copyright it to them > and add a note that I made contributions to it? Note it would be Copyright the code you and put a careful and detailed acknowledgement there. > impossible to go through it symbol by symbol and note all instances > where the code coincided with theirs. It just strongly resembles theirs > in those few functions, without being the same. > > 4) In the case of the NTL replacement library I am writing (NTL is > GPL'd) I can just contact Victor Shoup. In some cases I have virtually > just translated his code (from C+wrappered GMP to C+native GMP), so I > can probably make some of the code copyright Victor Shoup and just note > that I modified it. Such derived code would then be copyright by both you and Victor. > In other cases the algorithms and code are uniquely > mine. The whole package is of course heavily based on the structure of > NTL, so that is quite a different situation to 3 above. It is also much > easier to identify which functions are just translations of NTL > functions, and which have been completely rewritten by me, using > different algorithms and coding ideas. --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://sage.scipy.org/sage/ and http://modular.math.washington.edu/sage/ -~----------~----~----~----~------~----~------~--~---
