On Tue, 22 Sep 1998, Greg A. Woods wrote:

> [ On Tue, September 22, 1998 at 12:19:54 (-0700), Bill Studenmund wrote: ]
> > Subject: Re: More licensing flames...
> >
> > On Tue, 22 Sep 1998, Greg A. Woods wrote:
> > The thing though is that (L)GPL has a restrictive LICENSE on how you can
> > distribute such code.
> 
> Whenever I talk about copyright licenses I talk *only* about
> distribution or other acts that would contravene copyright law.
> 
> > Actually, sticking the program inside another program would be covered by
> > section 2 as I understand it. That's what GPL is trying to prevent. You're
> > right that the proprietary software house would still own all the other
> > parts. That's not the problem, though. They couldn't legally redistribute
> > the program if it included parts which were from GPL'd code UNLESS they
> > made the source for the whole thing available. 
> 
> Ah, no, that's exactly the opposite of what I've been trying to say.

It may be the opposite of what you're trying to say, but it is what
everyone else understands the GPL to be about.

> Remember, collective works are explicitly permitted under law to be
> covered by a copyright that is totally separate from the copyright that
> may be on the original component works.

Right. As I understand it, (L)GPL doesn't disagree. It just says that to
be able to stick covered code into a collective work, you have to
distribute the whole thing under the (L)GPL. If you don't want to do that,
then the author denies you permission to so redistribute the covered code.

> In the absence of case law that shows an interpretation stating
> otherwise, it is my understanding that copyright law, at least in
> Canada, will not allow the copyright on a component of a collection to
> affect the other components in the collection.  Only the copyright of
> the entire collection may be affected, *and* the only affect it could
> likely have would be to totally prevent distribution.  It is highly
> unlikely that the copyright license on a component of a collection would
> be able to force someone to reveal what is in effect a trade secret
> (eg. some other component in the collection), even if such a license
> made claims that doing so would permit legal distribution of the
> collection.

Why not? The company in question chose to use (L)GPL'd code. You/they
didn't have to.

[snip]

> Yes, but not in the way some claim the GNU GPL or the GNU LGPL (or
> indeed the Sleepycat DB copyright license) attempt to do.  Traditionally
> (eg. in the literary world) for a work to be include in another
> collection a one-time license must be granted to the author of the
> collection.  However there's little or no "tradition" for the case where
> a freely re-distributable software component is linked into another
> program, possibly one that's protected as a trade secret and where only
> binary versions of the resulting application are distributed with an
> "All rights reserved" collective copyright.
> 
> It is my understanding that under copyright law there's little or no
> distinction between a program being linked with a library and having
> that library distributed in a separate "file" on the same "media".  Both
> will likely be viewed simply as "collective works".  Although the GPL
> attempts to make this distinction on behalf of the law, I'm don't think
> it would wash clean in the legal arena.

I have a lawyer friend I'm seeing tomorrow. I'll ask.

[snip]

> What about a proprietary source code control system that merely invokes
> a GNU GPL'ed program (eg. GNU Diff, or GNU RCS, or some such)?
> 
> What about an operating system that relies heavily on GPL'ed components,
> such as the way NetBSD relies heavily on the GNU C Compiler?
> 
> What I'm trying to get at here is how one would make the distinction in
> the eyes of the law between actually linking library code into an
> application and simply accompanying an application with some other
> separate program.  How closely does one have to tie GNU GPL'ed code to
> an application before it is considered to be the same work vs. a
> separate work in a collection?

As I understand it, the FSF has said that what you do before you ship the
product is what matters. If you link a library into a program, (L)GPL
terms apply. If your program calls a GPL'd program at run time (like how
Matlab calls Ghostscript on Unix machines), as long as the GPL'd program
source is available, you're fine. Also, you can run time link GPL'd and
non-GPL'd code together. So you explicitly can make GPL'd lkm's which link
to *BSD kernels (and most likely *BSD klm's which load into a Linux
kernel).

So they've set boundries. You may be right that the boundries don't extend
as far as the FSF believes, but they have given limits.

Take care,

Bill