Picking a license

[Patrick Maupin]

On May 14, 10:20 am, Paul Boddie <p... at boddie.org.uk> wrote:
> On 14 Mai, 09:08, Carl Banks <pavlovevide... at gmail.com> wrote:
>
> > On May 13, 10:59 pm, Steven D'Aprano <st... at REMOVE-THIS-cybersource.com.au> wrote:
> > > On Thu, 13 May 2010 17:18:47 -0700, Carl Banks wrote:
> > > > 2. Reimplment the functionality seperately (*cough* PySide)
>
> > > Yes. So what? In what possible way is this an argument against the GPL?
>
> [...]
>
> > It's not.  It's an argument that the GPL doesn't do much good.
>
> Right. So nobody got the benefit from Qt under the GPL or PyQt under
> the GPL? Even the PySide developers seem hell-bent on picking over the
> work of the PyQt developers for ideas, although they obviously won't
> touch the code. Nokia seem to have accrued tremendous benefit from the
> existence of PyQt because I rather doubt that anyone would have
> bothered rolling a set of mature, usable Python bindings for Qt now
> had some not existed already and proved that dynamic languages are
> worth supporting.

Perhaps he should have said "the GPL doesn't do any more good than any
other commercial license."

After all, lots of software ideas proved their worth in proprietary
systems, and then were later cloned by FOSS developers.  In many
cases, these clones are, functionally, almost exact copies.  That's
why all the really proprietary people are hell-bent on trying to get
or maintain patent protection -- copyright doesn't protect
"inventions".

Would you have agreed had he had said that "MatLab's license doesn't
do much good" and assigned the same sort of meaning to that statement,
namely that the MatLab license prevented enough motivated people from
freely using MatLab in ways that were important to them?  Obviously,
it was important enough to enough people that they went and built the
GPLed Octave software, which now emulates MatLab very closely.  As I
think both Ed and I have said before, the GPL can be a great license
for a full-blown *program* (like Octave) that people can just download
and use, but is not always so great for program *pieces* that are
designed to be used in the programmatic equivalent of a "mash-up",
like PyQt/PySide, so the cloning of PyQt into PySide is as inevitable
as the cloning of MatLab into Octave, or Unix/Minix into Linux.  As
far as your comments about PyQt proving out the concept, well duh!
Just as there are a lot of proprietary programs that are relatively
useless and *won't* have any GPLed versions written, nobody's going to
waste time rewriting a marginally useful GPLed library just to put a
permissive license on it, either.  They are either going to write
something completely different in the hopes that their vision is
correct, or are going to copy at least some parts of the design and/or
vision of something that is popular and useful.  And that's a great
thing.  It would have been horribly unproductive if the Linux API
weren't at least reasonably close to the Unix API, for example.

It's an interesting tension between the licenses.  Nobody rewrites
permissively licensed software as GPL simply because the license is
unacceptable, but just as people will rewrite proprietary programs and
libraries as GPL because the license is unacceptable, so will some
rewrite GPLed libraries as permissive because the license is
unacceptable.

Regards,
Pat