[gnu.org #693634] GMP as dual-license GPLv2+/LGPLv3+?

[Karl Berry]

    simon> I'm not sure it made it there -- there wasn't any cc line in
    the e-mail copy I got,
     
That's RT being its usual helpful self.  (I'm cc-ing this message too.)

    and I can't find your post in the gmp-devel archive.  

It was held for approval.  (BTW, gmp people, I strongly recommend
respond_to_post_requests=no to eliminate backscatter.  Anyway.)

    input from the FSF

To repeat, I am not the FSF, but my thoughts are below anyway.

    niels> from a plain LGPLv3 library into a dual licensed GPLv2 &
    LGPLv3 library requires explicit permission from the
    author/copyright holder),

I agree that is a nontrivial drawback of the dual licensing.

    in the maintainers file is intended as a bugfix to LGPLv3 (that text
    could have provided use under "GPLv2 or later" as an additional
    permission) 

Personally I think that would have been a good idea; I have no idea what
rms would think or if the suggestion arose when (L)GPLv3 was being
drafted.  (And, I doubt he wants to think about an LGPLv3.1 at this
point.  I did suggest changing the LGPL instead of the dual licensing,
and he vehemently rejected that idea.)

    LGPLv3, or an exception to be used only when there's a compelling
    reason. If the dual licensing is going to be the rule or the
    exception

At any rate, given that it wasn't written into LGPLv3, the dual
licensing was intended as an exceptional case.  "Most" libraries, for
some definition of "most", don't need to support linking against
GPLv2-only applications (several have already upgraded).  In addition,
libraries that provide unique functionality should be GPL'd, not LGPL'd.

Overall, if you are happy enough with the dual licensing, then, as I
said, I am sure it is ok -- we came up with that stuff exactly for GNU
packages in this circumstance, after all.  If you feel unsure and want
to talk over the pros and cons with rms, feel free.

Best,
karl