[Coin-discuss] license issues

[Matthew Saltzman]

On Tue, 19 Sep 2006, Robert Fourer wrote:

> Since AMPL came up in the discussion, I thought I'd also mention a few 
> related issues ...
>
> It's true that the AMPL-Solver Library is only needed for building the 
> AMPL-interfaced version of IPOPT.  But to further complicate matters, 
> the AMPL-Solver Library has been released as open source, but under a 
> license different from either GPL or CPL (indeed COIN-OR was not yet on 
> the scene when the AMPL library was first made available).  I'm not sure 
> whether this adds further difficulties to the distribution of binaries 
> for AMPL/IPOPT, which would be convenient to have.
>
> AMPL only interacts with a solver by writing a file and setting some 
> environment variables, spawning the solver as a process, and at the end 
> retrieving a file written by the solver.  As a result I believe that the 
> licening issues surrounding an AMPL-interfaced solver are completely 
> independent from the issues regarding the core AMPL program (which is 
> not open-source).  It would be interesting to hear other opinions on 
> this, though.

For the AMPL solver itself, I think we are safe, as there is no actual 
linking going on.

For the library that the *solver* calls to read the input file and write 
the output file, we face all the issues that have been discussed here. 
My assessment is, there is probably no problem linking from CPL or LGPL 
code, unless the AMPL-Solver license makes an issue of it.  Neither of 
those licenses make claims on code that interacts with covered code 
through a well-defined API and linkage.

GPL code should be OK too as the AMPL-Solver library is the library and 
the GPL code is the calling code.  I don't believe there is anything in 
the GPL that would prevent, for example, compiling a GPL program with a 
commercial compiler and linking to a proprietary C Standard Library (as 
long as the library permitted redistribution in linked form).  At least 
I've never seen any discussion to that effect anywhere.  There's no way I 
can see to make the case that, as a general principle, the library being 
called could possibly be a derived work of the calling program.

(Of course, none of us are lawyers, etc., etc.)
-- 
 		Matthew Saltzman

Clemson University Math Sciences
mjs AT clemson DOT edu
http://www.math.clemson.edu/~mjs