Re: ocaml, QPL and the DFSG: QPL 6c argumentation.

[Sven Luther <sven.luther@wanadoo.fr>]

On Sat, Jul 24, 2004 at 09:29:36PM -0400, Brian Thomas Sniffen wrote:
> Edmund GRIMLEY EVANS <edmundo@rano.org> writes:
> 
> > Sven Luther <sven.luther@wanadoo.fr>:
> >
> >> > Do you think that the QPL without section 6 is a free software
> >> > licence?
> >> 
> >> I am tentatively in favor of that, yes.
> >
> >> > If YES, how do you argue that section 6 detracts from the permissions
> >> > granted by section 3?
> >> 
> >> They do not, since they apply to two different clases of software.
> >
> > That seems like a contradiction to me. You seem to be saying that the
> > QPL without section 6 is a free licence, section 6 does not detract
> > from the permissions granted by section 3, and yet we have to look at
> > section 6 in detail to tell whether the QPL is free. How does that
> > work?
> >
> >> What is your argumentation to ignore the above and makes as if modified work
> >> and linked works are one and the same thing ?
> >
> > It looks to me like section 6 grants some additional permission in the
> > case of mere linking. Without section 6 the entire work would have to
> > be QPL (with a licence grant to the initial developer). With section 6
> > only the part that contains the original software has to be QPL; the
> > rest can have any free licence, more or less, except that there's an
> > additional requirement (6c) that might be problematic.
> >
> > So the argument here is that the DFSG requires the conditions in QPL 3
> > to be acceptable, and if they are, then the DFSG is satisfied and we
> > don't have to look at QPL 6.
> 
> But we do have to look at QPL 6, because there's no way to say "We
> accept only the parts in 1-5, not 6; this is a modification under 3,
> not a linking under 6."  If I modify *and* distribute, I must obey
> both of those clauses.  Similarly, if I modify *and* link, I must obey
> both of those.  The QPL authors happened to impose extra restrictions
> on linking, and relax a few restrictions, but I think it's quite a
> stretch to read this the way you like.

Exactly, thanks Brian. That said, i dislike your "modify *and* link" approach,
since a given software can either be a modification of the original software
(which can replace it) or link with the original or modified software (and
thus use it). I have sever doubts that there exists a case where both clauses
happen. I believe that QPL 6 would also probably apply only to software
distributed in a separate tarball if you go down to it, but am not entirely
sure of it.

> In any case, linking is particularly important for OCaml.  The OCaml
> toplevel is under QPL 6, lumped in with the compiler.  Common practice
> is to distribute customized toplevels, which use particular extensions

Bah, no. That is hardly common practice. It is done for lablgl and both
lablgtk though, so you may have a point, and a reason for going upstream with
it, maybe asking for a derogation for it. Still both of these software are
DFSG free apart from the ocaml dependency that is under discussion, and widely
available, so this should not matter in practice.

> -- unix.cma, for example.  It's a common means of extending the
> language.

Yeah, but usually the users do it themselves, and not distributing the code.

So would we be satisfied if the maketop was excempted from QPL 6c ? I doubt
it, since you would then consider also separate files under the compiler which
are then reused in other software.

Friendly,

Sven Luther