RE: licencing and modules

> On Fri, Apr 20, 2001, Robert Collins wrote:
> > I think now is a good time to raise this topic.
> >
> > One of the concerns I have with getting run-time modules
> going, is that
> > corporations may release binary only modules. Thats
> something I don't
> > want to see happen. I'm driving the run-time loading to allow easy
> > inplace upgrades, and more flexible configs, not to reduce
> the GPLness
> > of squid.
> >
> > Is there any mechanism that can be put in place to ensure that all
> > libraries/modules that squid can load will have to be GPL'd
> as well?
>
> <flame>
> This hasn't stopped people in the past..
> </flame>
>
> Seriously though, does it matter? If companies start releasing
> binary-only modules for squid, I'll be flattered. And since I bet
> they'll need to fix up some internal bit of squid to get a little
> more performance or flexibility, the GPLness is still there.

Actually, dynamically-loaded modules, in RMS's explanation of
the GPL, form a "derived work" and thus MUST be GPL'ed as well.
This is a controversial matter, since it's not explicitly covered
in the GPL, but only in side-notes and declarations by Stallman
and others. For instance, Linus Torvalds
explicitly stated that in his opinion loading a Linux kernel module
does NOT constitute creating a derivative work.
But that has to be explicitly _allowed_ under the most common
interpretation of the GPL.
I suggest that we discuss the fact, and then explicitly add a statement
as an addendum to the copyright notices somewhere.

We might be "generous" and state that dynamically linking a module to
squid does not constitute a derivative work, or we might be Stallman-ish
and state that it does. Either way, I think that a statement will only
do good since it will remove this potential loophole either way.

I hope I've been confusing enough.

-- 
	/kinkie