Alex Bradbury via llvm-dev
2015-Oct-29 12:30 UTC
[llvm-dev] RFC: Improving license & patent issues in the LLVM community
On 29 October 2015 at 10:25, Jonas Maebe via llvm-dev <llvm-dev at lists.llvm.org> wrote:> Hi, > > Regarding the previously voiced concerns of incompatibilities between the > Apache and GPLv2 license, I'd like to add one more thing. > > I work on a, at this time mostly LLVM-unrelated [1], "GPLv2 or later" > licensed compiler: the Free Pascal Compiler. Some people in the project are > vehemently opposed to the GPLv3, so a move of our project to that license > would not be easy (even if legally perfectly possible). > > The reason I care about the LLVM license is that I regularly consult the > Clang source code to find out things that are not documented anywhere else. > In particular: > a) layout details of Objective-C class/category/interface metadata > b) OS X compilation details, e.g. for which versions of (Mac) OS X and iOS > do you have to link in which, if any, crt*.o file > > Not being able to use the Clang source code anymore as a reference point for > this kind of information would be a big loss to me. Apple just documenting > all of this would be equally fine by me, of course, but I don't see that > happening.(Disclaimer: I am not a lawyer). Copyright protects creative expression rather than ideas or facts (although some jurisdictions do have a copyright-like idea of 'database rights' https://en.wikipedia.org/wiki/Sui_generis_database_right). Referring to Clang source to ascertain certain facts wouldn't necessarily put any requirement on the licensing of the code you later produce. However for safety many might prefer not to look at the source so that any incidental similarities can't be accused of being due to copying code directly. If you're worried about this, asking a community member to use the Clang source to write a short document on the details you require might be a way forward. Alex
Jonas Maebe via llvm-dev
2015-Oct-29 12:54 UTC
[llvm-dev] RFC: Improving license & patent issues in the LLVM community
Alex Bradbury wrote on Thu, 29 Oct 2015:> However for safety many might prefer not to look at the source so that > any incidental similarities can't be accused of being due to copying > code directly. If you're worried about this, asking a community member > to use the Clang source to write a short document on the details you > require might be a way forward.Definitely, but so would LLVM retaining a GPLv2-compatible license :) And it definitely would be much easier than finding a Free Pascal community member with spare time that knows enough about both C++ and compiler development to be able to carry out that task (our compiler is self-hosted/written in Object Pascal). Maybe just dual licensing everything under both the Apache 2 license and a GPLv2-compatible one (maybe even the GPLv2 itself) would be a solution? I don't remember whether this was already considered/discussed in this thread. Jonas
C Bergström via llvm-dev
2015-Oct-29 13:34 UTC
[llvm-dev] RFC: Improving license & patent issues in the LLVM community
I wish this license thread would die. I'm really disappointed that Chris would start a mess like this. I hope in the end it gives fruit to something positive and tangible. ---------------- In terms of GPLv2 + APL2 compatibility. All we have is the views and opinions of various sides. Nothing has been decided in court (afaik). Until someone actually litigates this - we just have FUD from GNU. Their opinions are probably biased more to scare people into GPLv3 than based on something they could win. If you really feel strongly about this - maybe email SFLC or take it to a more appropriate channel (offlist) or not here.
Joachim Durchholz via llvm-dev
2015-Oct-30 13:20 UTC
[llvm-dev] RFC: Improving license & patent issues in the LLVM community
Am 29.10.2015 um 13:30 schrieb Alex Bradbury via llvm-dev:> However for safety many might prefer not to look at the sourceThose who do that have no interest in the LLVM license anyway. IOW I do not think that this needs to be considered in licensing choices. (Correct me if I'm wrong.) > so that> any incidental similarities can't be accused of being due to copying > code directly.From what I see in the SVN, FreePascal is written in Pascal, so accusations of verbatim copying from the LLVM sources wouldn't work. That's still not 100% safe, but then I guess nothing is. Regards, Jo
Jonas Maebe via llvm-dev
2015-Oct-30 13:44 UTC
[llvm-dev] RFC: Improving license & patent issues in the LLVM community
Joachim Durchholz via llvm-dev wrote on Fri, 30 Oct 2015:> Am 29.10.2015 um 13:30 schrieb Alex Bradbury via llvm-dev: >> so that >> any incidental similarities can't be accused of being due to copying >> code directly. > > From what I see in the SVN, FreePascal is written in Pascal, so > accusations of verbatim copying from the LLVM sources wouldn't work.A "derivative work" in copyright is much broader than verbatim copying. Plain translation into another language also falls under it (try translating a Harry Potter book and publishing it yourself :) Jonas