[License-discuss] GPL and proprietary WebAPIs

[Chris Travers]

On Fri, Dec 23, 2011 at 9:49 AM, Chad Perrin <perrin at apotheon.com> wrote:
> On Fri, Dec 23, 2011 at 03:38:04AM -0800, Chris Travers wrote:
>>
>> Thus in general I think one is generally better off talking with
>> upstream projects and trying to get them on board.
>
> Take the most restrictive reasonable interpretation of both if you want
> to play it safe.  After all, a change in the upstream project's
> maintainership could get you in a lot of trouble if you rely entirely on
> the legally non-binding word of a project maintainer.

I think one could easily read the GPL v3 and the 2-clause BSD licenses
(and hence every other similar permissive license) as requiring
incompatible things.  If we want to take the most restrictive
reasonable interpretation of both, these licenses are incompatible,
which is somebody nobody really believes.

The issue comes down to this:  The general view among
BSD-license-using developers is that you cannot change the license of
a BSD-licensed work as such without changing it (and then you are
creating a new work and effectively licensing your own work under
terms you choose provided that required notices are intact).  In other
words, the predominant view of the BSD license among developers who
license their works in that license is that the license to the code
does not restrict your own ability to license your own code which may
use this code however you wish, but the license cannot be changed on
the code itself.  This view is also expressed in the SFLC's discussion
of mixing permissive licensed files with GPL'd files and advises
*against* changing the license without adding code which is at least
potentially copyright-worthy.  This view is clearly reasonable.

The view of RMS and Eben Moglen (at least last time I discussed this
with Moglen, albeit a couple years ago) was that the GPL v3 *requires*
the ability to change the license without actually adding code
(section 7, paragraph 2), but that such terms can't effectively
enforced without changing the work.  In other words, GPL v3 according
to them requires that the license be convertable to the GPL v3 without
adding new code.  Many people feel this view is reasonable including
many on this list.

In other words, the BSD license cannot be assumed to be reducible to
the GPL v3 + additional permissions as defined in section 7 of that
license.  For the licenses to be compatible, the GPL v3 must be read
with the assumption that they are compatible (this is one area every
lawyer I have talked with on this matter agrees with), and so you have
the question as far as whether the 7(b) additional restrictions to
preserve legal notices are a sufficiently broad area to essentially
provide additional permissions which cannot be removed by someone who
merely distributes the source of the software.  If it is, then these
licenses are clearly compatible.  If not.....

In other words if you take the most restrictive reasonable
interpretation of each license, then we can't have GPL v3 software
using standard PostgreSQL client libraries for example (because you
can't just relicense these libraries as under the GPL without adding
copyright-worthy changes).  I want to see anyone who really agrees
with this.

The fundamental problem here is something that lawyers try to
eliminate but fundamentally cannot:  human language is fundamentally
subject to ambiguity and for every text there are a significant number
of potential interpretations of it.

Best Wishes,
Chris Travers