[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: GPL "+" question

Paul Tagliamonte writes ("Re: GPL "+" question"):
> They *can* since the work as modified *can* be distributed under the
> terms of the GPLv3+, *without* changing the original work's license, but
> the *file* can be distributed as GPLv3+, since that's the minimum
> license needed to comply with all parts.

I think the problem here is the notion that a file necessarily has
exactly one licence.

The only actually accurate statements one can make are "there is [not]
permission to distribute file F under licence L".

If you want to know "what licence does a file have" what you really
ean is "for what set of licences L1,L2,L3... is it the case that there
is permission to distribute".

So it is true that a downstream redistributor who does not change F
cannot "change the licence", because the only permission needed (in
copyright law) is that from upstream.  If upstream grant GPLv2+ then
even if the downstream writes GPLv3+ then everyone can still rely on
upstream's permission, putting the upstream GPLv2+ notice back.

However, if a downstream redistributor modifies the file, they can
narrow the set of permissions.  This is because in copyright law, they
have their own copyright in the modifications.  So if a downstream
changes GPLv2+ to GPLv3+ _and modifies the file_ then the GPLv2+ is no
longer applicable.

For Debian, there is the question of what to put in debian/copyright.

Obviously we need to put in debian/copyright some applicable Ln (that
is, an Ln which applies to all the files F).

If there is permission to distribute under L1,L2,... then it would be
legitimate in a copyright law sense to write in debian/copyright only
L1.  To do so would not be a breach of the copyright, because we would
be acting in accordance with L1.

(So I think Paul is wrong if he thinks, as he seems to, that it is a
violation of copyright law to change GPLv2+ to GPLv3+ when merely

But to do so would be rude and we should try not to do it.  So I think
that where practical if we can determine that a package is dual
(or triple) licenced, we should document all the permissions - that
is, all the licences.

It appears that in this particular case package has been modified by
the GPLv3+-preferring downstream.  In that case there is no permission
to distribute under GPLv2+ any more.

I don't think it is sensible to insist that the Debian maintainer do a
lot of work to try to discover whether some files in the package have
/not/ been modified by the GPLv3+-preferring author (or by other
people who contributed to that author's version and never saw a
GPLv2+ licence).

That work is not necessary in copyright law, and we aren't really
doing the Free Software world much of a service by performing it - at
least, once we have decided to package the GPLv3+ fork at all.

If someone wants to try to strip the GPLv3+ parts out of the fork
then that's fine of course, but I don't think we should insist that
the Debian maintainer do the necessary archaeology.

> The proof is on you -- where does it say you can relicense someone
> else's copyrighted work / IP? Not *redistribute*, *relicense*.

The answer to this question is very clear.  The proof is in the
original licensing notice:

  either version 2 of the License, or (at your option) any later

The downstream distributor may choose a later version (3, say, or "any
version 3 or later) and comply with its terms.


Reply to: