RE: Linus about GPL x BSD(long)

On Sat, 20 Jul 2002, Patrik Stridvall wrote:

> > You don't believe the GPL's viral mechanism would stand a court 
> > challenge?
> 
> For the most part, no.

I guess we'll just have to see when a test case eventually comes along.

As I understand it, the GPL is part copyright license, part contract -- the 
contract grants the license, and if you don't agree to the contract, there 
is no license.  Therefore, in theory, copying it is only legal because of 
that contract, and without the contract, copying is illegal.  (But I'm not 
a lawyer, so I'm just judging from the explanations I've heard.)

Of course, a sticky point is that contracts normally require consideration, 
and since the author doesn't receive anything of tangible value, it might 
not be considered a valid contract.  (Perhaps this varies from one state to 
another.)  It would be rather bizarre if the contract is ruled invalid in 
general, and everyone following the GPL and using it as intended suddenly 
becomes guilty of copyright infringement!  (Hey, it could happen...)

> I think the main issue is do you distribute the GPL:ed work in whole
> or in part at all.

If it's only in part, it's possible that it could fall under "fair use" and 
be distributable without a license.  If it's a small enough part...

> Whether your code is meant to statically link or dynamically link or
> interact with the GPL:ed work some other way is IMHO opinion irrelevant.
> They are all the same thing from an abstract point of view.

Your source code is no more or less derived from the GPL code, regardless 
of how it's linked.  I agree with that.

> You will have to avoid that your code becomes a derived work of course.
> However that will IMHO in most cases be quite easy so that wouldn't be
> really be problem.

The FSF claims if you program to a unique API that only exists in GPL code, 
that makes your code "derived" from the GPL code.  However, I don't see how 
this is conceptually any different from a clean-room implementation where 
someone has described the API, and someone else codes to that description 
without reference to the GPL code.  Somehow I doubt this theory of source 
code derivation due to API usage would hold up in court...

> As I said, I think that as long as you doesn't actually distribute
> the GPL work itself you are pretty safe. Just do as FreeBSD does
> in it's port collection. Just provide an URL and a script that the
> user can run so that it download and compiles everything.

Sure, that should be as safe as dynamic linking, and leaving out the GPL 
shared libraries from the distribution.  (Which isn't safe at all, if the 
FSF's API-derivation theory holds...)

My point was that distributing a static binary is clearly distributing the 
GPL code, since it has been incorporated into the binary.  If you choose to 
distribute a static binary, you have to abide by the GPL to benefit from 
the copyright license in it -- then the contract provisions in the GPL (if 
it's a valid contract) would obligate you to release your source -- even if 
your sources aren't inherently "derived works" in a legal sense...

> If you actually distribute that GPL:ed work is some form that is
> quite true. I wouldn't want to do that until the court clear this up.

Well, if you're willing to follow the GPL's rules, it should be safe enough 
to distribute source and binary.  If you're trying to keep your source as 
closed as possible, you ought to avoid the GPL entirely until it's actually 
been tested in court.

> But then, with the Internet there is very little need to actually do that.
> The user can do that himself with the help of appropriate scripts...

Yes, this is easy enough.  Of course, the FSF has always contended that 
"user does the link" doesn't excuse you from the obligations of the GPL.  
(This was one of the first ideas on how to circumvent the GPL, after all.)

Whether or not a court of law would agree is an open question.

Deven