Re: Lilypond \include statements and the GPL

[Alexander Kobel]

On 04/02/2013 10:04 AM, Joseph Rushton Wakeling wrote:
> On 03/30/2013 01:02 AM, Alexander Kobel wrote:
> > On the other hand, user C /should/ be allowed to distribute source code under
> > whatever license he wants to /as long as he doesn't ship the GPL libraries with
> > it./  It's useless without them, but anybody who wants to run or compile the
> > code is free to download the necessary GPL'ed stuff.
>
> If I write a computer program which uses functions from a GPL'd library, it
> doesn't matter whether I distribute an executable or just source code, and it
> doesn't matter whether I distribute the source code alongside the GPL'd
> libraries or as an individual file.  It's a derivative work under the GPL and
> must be licensed accordingly when distributed.
>
> See: https://www.gnu.org/licenses/gpl-faq.html#IfInterpreterIsGPL

This certainly applies to compiled code, with the GPL'ed library 
statically linked, and also (I stand corrected) with dynamic linkage, 
AFAIU.  I still cannot see how it /could/ possibly apply to source code:

Case 1: If I write a C or C++ application which needs, say, some numeric 
functionality, I can get it from Eigen, Intel MKL, ATLAS, CUDA-based 
BLAS, ....  (I don't know the exact details, but I'm pretty sure that 
some basic functionality can be achieved with the same syntax for all of 
them.)  In that case, it's not even clear /which/ library the code 
/could/ be linked against (just assume that the headers included are not 
referenced by library name, but via a pre-compile-time dispatcher or 
user input at compile time).
Case 2: If I want to use CGAL, I have two options to get the library: 
GPL/LGPL'ed or under an industrial license.  The code looks the same - 
just that the commercial license definitively allows me to distribute 
the code.  But once it's distributed, it cannot be distinguished from 
the same source code, using the free license.  Does that mean that 
whenever someone looks at that code and used GPL'ed CGAL, my code 
suddenly goes GPL???
Case 3: Assume I reimplement some API and functionality of a non-open 
library as a GPL'ed work.  Again, whether the application uses the 
former or the latter library cannot be distinguished until somebody 
compiles with one of them.

In that sense, I have the feeling that the source code license cannot 
interfere with the library license, unless parts of the library are 
copied in the source code.  The source might be useless without the 
library, but still.  E.g., is it really the case that the boost project 
would never be allowed to include wrappers to GPL'ed libraries?

(Sorry for the strange examples; I first thought about something like 
the STL, but this qualifies as a system library, which - for good reason 
- has an exception.)

> Perhaps you are thinking of e.g. the case with UNIX shell scripts, where e.g. I
> can write a script that calls GNU sed without having to license my script under
> the GPL.  But this is because there's no dynamic linking that takes place when I
> do so -- I'm starting an independent process and receiving its output.
>
> See e.g.: https://www.gnu.org/licenses/gpl-faq.html#GPLAndPlugins

I see your point, but I don't see how what you explained to be the GPL 
notion could actually be enforced...


Best,
Alexander