> I consider this the absolutely worst reason to use the GPL.
I think rms is right here. Selling exceptions is a good thing. It's certainly no worse than selling proprietary forks of non-copyleft free software:
http://www.gnu.org/philosophy/selling-exceptions.html
Why would it be ok to take non-copyleft software like clang and make proprietary forks with it, but not ok to sell exceptions in order to fund copyleft software?
EDIT: As an Octave developer, I greatly respect how FFTW's copyleft is used, for example. Matlab buys a license exception to use FFTW. Octave gets FFTW under the terms of the GPLv3, just like Octave itself, which we in turn contribute back to everyone else. Matlab and Octave are both using the exact same library for Fourier tranforms.
I like this model. Everyone is paying one way or another, no freeloaders, just lots of freedom.
I think the problem I have with the practice of selling exceptions is that it gives more rights to the owner. I feel that open source licenses should give equal rights to everyone. Dual licenses undermine that equality. I understand that people want to stay in control of their software, but for me the words 'free' and 'open' somehow imply 'equal rights for everyone'.
I'll ask the FFTW guys what do they do about external contributions.
Also, thinking about it more, there is nothing wrong with copyright holders having extra rights as long as they do not use these rights abusively. After all, they did the work, so they deserve to be compensated for it. Nobody would dispute that. It is problematic if they use these extra rights to try to make people pay for proprietary software that is not otherwise available freely. This is what I dislike about what Oracle does.
The problem is that many Open Source projects aren't created by an individual, or even a single company. Large projects have hundreds of contributors. If a project requires copyright assignment, all these contributors have fewer rights than the maintainer of the project.
If a project doesn't require coypright assignment, and uses a single license, all contributors have equal rights to the collective work.
It's some sort of empirical law that almost all of the code in free projects is written by at most a handful of all contributors. If this law holds, and I can think of only few examples where it doesn't, it seem fair for the tiny minority to benefit the most from selling exceptions.
What the FSF does when requesting copyright is a two-way agreement. The FSF drafts a contract that promises that they will never use the copyright aassignment for creating non-free software. Similar contracts could be drafted that agree to share revenue proportionally and provides clauses for arbitration if this proportion is in dispute.
None. But if they do that then you can fork the latest free release and carry on improving that (e.g. OpenOffice -> LibreOffice), and from then on the company's version receives no more contributions from the community.
And it's unambiguously better for the principles of the GPL than the alternative of licensing the project under a BSD-style license and requiring contributions to be made under such a license, which allows not only the original maintainer but anybody whatsoever to do the thing you're objecting to.
I consider this the absolutely worst reason to use the GPL. Quite a few companies (large companies like MySQL/now Oracle or small ones like Macromates) require contributors to sign an agreement that allows the company to distribute the code under a different, proprietary license.
Doesn't this circumvent everything the GPL stands for? If I contribute to such a project, what guarantee do I have that the owner doesn't just distribute the software, including my contributions, under a proprietary license and stops making improvements public?