Re: GPL V2 and GPLv3
Peter S Galbraith writes:
>> > If it explicitely calls the interface...
>> ... that does not resolve the question, even for bytecompiled files.
> So why do need the freer LGPL versus the GPL for libraries? I guess we
> can do whatever we want so it doesn't matter.
When there is only one implementation of a library, there is an
argument -- not everyone agrees with this, but I will assume it is
strong enough to be cautious -- that the inclusion by reference is
enough to make the library's license apply.
That is not the case for Emacs, especially for standalone packages
that are not distributed as part of an upstream Emacs package: the
same elisp file typically works with more than one version or branch
of Emacs (XEmacs is still alive and is still GPLv2 or GPLv2-or-later).
>>>> In this case, there are older emacsen -- distributed under licenses
>>>> other than the GPLv3 -- that provide the interfaces needed by most or
>>>> all of the elisp in question.
>>> Sure, the code is fine with older Emacs. We simply shouldn't install it
>>> and set it up for GPL v3 versions of Emacs.
>>I suggest you take this question up with the FSF. If their position
>>is that there is a license conflict between GPLv2-only elisp files and
>>a GPLv3 (or later) installation of GNU Emacs, it is worth talking
>>about -- in part because there would be wider-ranging implications.
> The FSF have stated that the two licenses are not compatible.
That's a different question than what I asked.
> In the past, I had asked RMS if it was reasonable that elisp code could
> be under any license if it simply used Emacs builtins, and would have to
> be under the GPL if it pulled in GPL'ed libraries using a `load' or
> `require'. He agreed that it made sense, because the libraries weren't
> under the LGPL but under the GPL. Transpose that to the incompatible v2
> or v3 and you get the same situation.
You have not listed any GPLv2-only examples that pull in GPLv3-only
libraries. Those would be libraries that do not exist, or which have
versions that would break the GPLv2-only elisp code's function, in
GPLv2-compatible versions of Emacs. If you know of any examples,
please mention them.
(This is one of the reasons that -legal posters so often say that the
DFSG apply to software, not to licenses: there are an almost infinite
number of hypothetical situations, and there is neither the need nor
the time to evaluate most of them.)
Otherwise, the only way I see any problem is if you think there is
GPLv2-only elisp, valid to distribute with prior versions of GNU Emacs
(and/or XEmacs), but that somehow became illegal to distribute when
GNU Emacs was released under GPLv3.
>>So far, though, you seem to be the only one with that opinion, and you
>>have not provided any support for it beyond assertion.
> Wow, I didn't think it was controversial thinking. I'm simply seeking
> advice here, not trying to change the world.
> The point for me isn't the letter of law but the intent of the license
> and the code's author. I can understand perfectly if someone didn't
> want to give a blank cheque and specify "GPL version 2 or any later
> version" for their code because they'd rather decide that when the new
> version actually comes out. By that token, code that isn't compatibly
> licensed won't be enabled "by me" in the packages that I maintain,
> because I consider elisp files as GPL'ed libraries. You can chose to do
> whatever you want.
> Hopes this helps to clarify where I come from.
The meaning of the law is very relevant -- GPLv2 makes explicit
reference to legal terms of art (such as "derivative work"). I do not
see any reason that GPLv2-licensed elisp has an inherent license
conflict with a GPLv3-licensed Emacs. Rather, section 0 of GPLv2
seems clear to me that this would be a permitted use of the code.
Even when the elisp is distributed in byte-compiled form (which is not
the case for standalone Debian packages), section 3's "mere
aggregation" language seems to apply.
It is polite and advisable to follow the author's wishes to the extent
that doing so is practical, but right now I see no reason to think
that any elisp author wishes his GPLv2-only elisp to not be used with
a GPLv3-licensed Emacs.