GMP license problem, anyone?

[Bill Hart]

On 29/05/2008, Vincent Lefevre <vincent at vinc17.org> wrote:
> On 2008-05-29 07:53:45 +0100, Bill Hart wrote:
>  > Firstly FLINT actually uses extensive portions of GMP in its integer
>  > library. It doesn't just link against GMP, though it does that also.
>  > If I want to keep doing that, I am limited to what was available in
>  > GMP 4.2.1 if I want my project to be part of SAGE, which is at least
>  > philosophically committed to GPL v2+. SAGE doesn't just link against
>  > FLINT, it includes it, the same as it includes GMP.
>
>
> If I understand correctly, you could still release some parts of the
>  project under (L)GPLv2+ and other parts (what is related to GMP 4.2.2)
>  under LGPLv3+. When you build and link the whole, it will be under
>  GPLv3+, but is this a problem? Why?

Probably you meant to quote the other paragraph, for which this may
appear like it is a solution.

But there are lots of problems with it. It doesn't work for SAGE,
which aims to have a fully v2+ distribution which includes the whole
of FLINT.

It doesn't work for any other project committed to v2+ which may wish
to include all or even a large portion of FLINT.

The next generation FLINT will have an integer library for which each
function is basically half from GMP and half my own code. But what is
the use of half a function!? Also it's annoying to document license
changes mid function. It's even worse when you consider that some
functions in GMP will be v2+ and some v3+.

But most important of all, it doesn't help me in that I don't want to
promote the use of version 3+ licences, since the problem just
proliferates. I want whatever I release, whether a function, a module
or an entire project to be fully compatible with GPL v2+.

Bill.