On 2009-09-01 at 13:52:16 [+0200], Bruno Albuquerque <bga@xxxxxxxxxxxxx> wrote: > >>> I'd love to hear the interpretation from someone who is very much in > >>> the know, but who is not also a GPL zealot... > >> > >> And this is the reason for my meeting on Monday. Daniel Berlin is > >> actually a lawyer specialized in open source. Chris DiBona is > >> responsible for making sure we conform to open-source licenses at > >> Google. > > > > I was wondering... How did the meeting go? > > Oh, sorry. Completely forgot about this. The meeting was with Chris only > as Daniel didn't make it. His opinion (and I have to stress opinion here. > This can be in no way constructed as legal advice) is basically what we > already concluded. The GPL is basically so convoluted that the only way > to play it safe is to, when in doubt, avoid it if at all possible (unless > your project is GPL, of course). When it is not possible, contacting the > original authors may help but it could be difficult with projects like > ffmpeg as there are dozens (if not hundreds) of contributors so unless > the copyright is signed to an entity that has a legal right to make an > exemption in an specific case, this will not help either. We should be on the safe side now. The alpha-builds are not configured to include GPL-addons. In this mode, FFmpeg does not include GPL code, it's completely LGPL. We do not include the GPL AC-3 decoder, but instead have upgraded FFmpeg which now includes an LGPL AC-3 decoder. The file-systems, for which the FSF GPL-FAQ mentions an exemtion, now ignore the GPL configuration and are installed anyway. While I don't agree that the GPL-FAQ has more say than the GPL itself, I am personally fine with this setup. This should mean we are fine as far as the GPL issue is concerned. The other question is about possibly patent-violating code we include. We are disabling it for Freetype, but are still including FFmpeg, which "probably" violates patents. The thing is of course that a lot of other code in our repo "may potentially" violate patents, given the number of patents out there and nobody having done any research to avoid them. We can guess that something probably violates patents, but have no resources to research any of this. So the status quo is that we disabled the code that we directly know would violate patents... is everybody fine with this? Personally, I could live with that, especially since software patents are not yet valid in Germany. The question is whether we should even go as far and re-enable the Freetype code. Some time back I read an article on the subject, and at least the subpixel/Cleartype patent is most likely not valid, let alone we do not use the filtering from Freetype, but do our own totaly trivial filtering. If at all, I would say only the Truetype bytecode-based hinting is worth being concerned about. Best regards, -Stephan