Questions / Concepts GPL. Was: Re: GPL License with clause for Web use?

simo simo.sorce at
Fri Nov 30 15:25:46 UTC 2007

On Fri, 2007-11-30 at 09:39 +0000, Sam Liddicott wrote:
> * simo wrote, On 29/11/07 22:57: 
> > > It's how I read the AGPL - except the GPL3/13 allows conversion of the
> > > GPL3 work to AGPL in which case the modifications could be AGPL.
> > > 
> > > It seems to me that GPL3/13 that gives authority to do this, not AGP -
> > > if the recipient chooses to act on that authority and do so.
> > > 
> > > However, a summary of points will be collated on Monday for a request
> > > of official clarification as per Shane's suggestion.
> > > 
> > > I enjoyed the discussion, anyway... even though I got a little
> > > overheated, so thanks all, for being so frank so that we could
> > > understand each-others concerns and understandings.
> > >     
> > 
> > 
> > I have exchanged a couple of emails with Richard Fontana of SFLC, one of
> > they key lawyers involved in helping FSF coming up with the GPLv3 during
> > the drafting process.
> > 
> > He told me his personal opinion, which is not necessarily what the FSF
> > thinks, is that my read of the clause is close to his reading.
> > 
> > Basically, Richard thinks that the "linked or combined" language does
> > not imply modification. IE, releasing a patch against the GPLv3 part of
> > the work under AGPLv3 would even be a copyright violation. The patch
> > needs to be GPLv3. Only the combination of the works obeys to AGPLv3's
> > additional requirements. But each piece retains completely its license.
> > 
> > Therefore there is no risk that a GPLv3 work can be effectively turned
> > into an AGPLv3 work by means of patches.
> >   
> Thanks for following this up.
> I hope then, that "combine" does not permit modification.
> (Please don't think I am being pedantic here, I try to understand the
> scope of rules by discovering their boundaries, and that is what I am
> doing here, and I appreciate your indulgence and help.)

Yes I think "combine" does not include modification, just those cases
where "linking" was not enough to include "potentially license
infringing" interactions beyond linking.

> However, if the modifier can't release an AGPL patch to the GPL3
> portion they may choose to not release a patch, but to merely continue
> to distribute the combined work as permitted by the AGPL.

No, I think you can *not* release that patch under the AGPL. That patch
will always be GPLv3. When you let people download the source code it
will be on the GPLv3 work side and clearly GPLv3.

>  It seems like the AGPL then would prohibit the separation of the
> patch from the combined work (which may not be what the license
> authors intended)

I don't think this is possible or make sense to me.

> Do you think from what Richard says that the original GPL3 author can
> extract this modification to the GPL3 portion as a GPL3 patch, (the
> patch not being his work and only distributed as part of the AGPL
> combination) and combine this patch with his own work and distribute
> it according to the GPL3?

I think, and Richard opinion seem to be the same, that you simply can't
release that patch "under the AGPL". All you do is release it under the
GPL together with the rest of the GPL work. Only if the application is a
combination of AGPL and GPL code, then you have to provide for download
also the GPL part. But the GPL stuff is *always* GPL, it is never AGPL.

> An example case: if a glue-class is written in php, subclassing a
> class from the GPL3 module, but instantiated by the AGPL3 module (thus
> also performing a link operation), and it be this subclass that has
> some implementation that "ought" to be incorporated into the GPL3
> module, will the GPL3 author, receiving the source by virtue of being
> a "user" of the AGPL3 work, be able to combine parts of this subclass
> into his own work, without receiving addition permissions fro the
> subclass author?

I think the "instantiation" is the linking/combination, but the class is
tied to the GPLv3 work. Therefore, my interpretation in this case, would
be that the class is GPLv3 bound. But this is all thin air, I think we
need to see a very specific case to determine that with some clue.

I think in some cases it can be very hard, but I guess it all boils down
to the authors defining what they consider being the boundaries.
If I were an AGPL author I would think twice the way I link to GPLv3
stuff and err on the GPLv3 side for any glue code.


More information about the Discussion mailing list