Re: GPLv2 licensing issues

[Yavor Doganov]

В Mon, 14 Apr 2008 14:32:43 -0400, Hubert Chathi написа:

> I don't think that the combined work still violations LGPLv3, because
> section 4 of the LGPLv3 allows you to release the combined works under
> any license that you choose, provided that you do certain things, and
> the library itself can still be distributed under the LGPLv3.  The
> exception allows the library to be distributed under the LGPLv3, so it
> should be fine AFAICT.

I believe that GPL'ed applications linking with LGPL'ed libraries 
actually use the libraries under GPL (using the direct upgrade clause, 
which is basically what makes LGPL fully compatible with GPL).

That is why it is possible to link a GPLv3 program with LGPLv2.1 library, 
because GPLv3 is one of the upgrade options of LGPLv2.1, and not because 
the GPLv3 program must follow section 6 of LGPLv2.1.

The reverse (our case) is not possible, since LGPLv3 talks about 
installation instructions and other things which are not present in 
GPLv2.  It might be possible to allow such combination with exceptions 
both for the app's and the library's license, but it won't help at all 
for Terminal and PopplerKit.  Also, such an exception to GNUstep is 
likely to wipe out the benefits of LGPLv3 and render the license upgrade 
more or less moot.

OTOH, if the copyright holders of the GPLv2-only apps are available to 
relicense their apps to "GPLv2 only plus (IMVHO rather dubious) 
exception", they may as well just upgrade the license.  The problem we 
have is for apps that cannot be relicensed, and no gymnastics with 
GNUstep's license is likely to solve that, except downgrading to LGPLv2.1.

(Little bit off-topic:
When Emacs was about to be upgraded to GPLv3, RMS asked to check all 
libraries it links against.  The compatibility matrix was not published 
(I think) back then, and this compatibility matter was even fuzzier.  
There was a concern about cairo, which was under LGPLv2 only [1], but 
newer versions are under LGPLv2.1 only (later it was found out that there 
is no problem with LGPLv2 as well).  Notice [2] which clearly shows which 
criterion applies when considering whether the result of the combination 
is distributable.

[1] http://article.gmane.org/gmane.emacs.devel/74038
[2] http://article.gmane.org/gmane.emacs.devel/74065
)