Il 16/05/2012 19:06, Kai Meyer ha scritto: > 1) It's been suggested to me that since we have the rights to distribute > our closed source shared library, there is a precedence for being able > to distributed a modified version of qemu that does run-time linking > against our shared library. The absence or presence of our shared > library simply enables or disables support for our file format. We are > happy to make available all changes to the qemu source code, but we are > not in a position to re-license our shared library's source code to a > compatible GPL license. This seems to be in contradiction to Paolo's > statement above, so while I can't resist asking if this is possible, I > don't have any realistic expectation that this is acceptable.
That's really getting into grey areas. IANAL, so I cannot answer this question. But as an aside, the GPL _does_ give you rights to distribute any modification you make to QEMU. The right questions to ask are: 1) a practical question: would the QEMU community accept that contribution? The answer here is "most likely not". 2) a legal question (i.e. the question that a court would answer): what are the rights of the _recipients_ of your version (and especially of the copyright holders of QEMU)? Do they have rights to ask you for the source code to the shared library, and to receive it under the GPL? Again I cannot answer here (and I couldn't even if I were a lawyer). (Remember that however what we proposed was not just relicensing your library source code, but alternatively to rewrite it from scratch with no particular attention to performance. That would be a completely different story, probably also for your lawyers. You could also share any internal spec you have and hire someone to write the QEMU interface for you, basically a form of clean-room reverse engineering). > 2) The GPL has provisions for you to create an exception where you have > specified a controlled interface. Am I right that qemu has not added > this controlled interface exception for file format access? What are > your thoughts on adding this exception if it is not present? I would > think that "struct BlockDriver" would make an excellent candidate for this. This would have to be applied to all files (not just block/*.c say) and agreed upon by all QEMU copyright holders. The second condition is quite obvious, the first I'll spend a few more words on. The first condition is because the code overall can be distributed as long as it fulfills all existing licenses. QEMU right now has files under BSD, GPLv2, GPLv2-or-later, LGPLv2.1-or-later and perhaps some more licenses. You can take code from individual files (or complete files) and reuse it under the license indicated in the header of that file. However, you can only distribute QEMU as a whole under the intersection of those licenses, which is GPLv2. If you add another license to the mix ("GPL+controlled interface") for block/*.c, QEMU as a whole could still only be distributed under GPLv2. > On a personal note, I am an open source enthusiast, so the last thing I > would want to do is to help alienate the relationship between qemu and > storagecraft. I'm not asking these questions to look for a legal corner > to worm my way into, but because I love open source software, and I want > to learn how to play nicely. (Plus there's that virtualization > "coolness" factor to this solution that I can't resist.) Sure, personally I appreciate your honesty even though I disagree with your goal. :) Paolo