On Sun, Aug 24, 2003 at 12:58:32PM +0100, Andrew Suffield wrote:
> > > From this I can conclude that I need to be able to distribute any
> > > given component of the glibc source code under the GPL.
> > Which isn't correct.
> > You need to be able to distribute the end product under the terms of
> > the GPL, which you can; the original parts don't matter, since if you're
> > distributing those, you're not bound by the GPL at all.
> Consider if a package consisted of two files. One is 'hello.c', and
> the other is 'fsf-funding'. The license for hello.c says "You may only
> distribute this along with a copy of fsf-funding, md5sum
> <value>". The license for the package as a whole says:
> * This program is free software; you can redistribute it and/or modify
> * it under the terms of the GNU General Public License as published by
> * the Free Software Foundation; either version 2 of the License, or
> * (at your option) any later version.
> The licensor of the package does not own the copyright to hello.c,
> and therefore cannot be construed as implicitly licensing it under the
> GPL.
If the package as a whole just includes hello.c and fsf-funding, then
there's no independent material to put a copyright on, and trying to
put the GPL on it is meaningless. But let's go with it, and assume that's
not the case.
> You can "distribute the end product under the GPL". You cannot remove
> fsf-funding and distribute the result.
But that means you can't distribute the end product under the terms of
the GPL, which include (in part 2) the ability to make modifications
only taking into account a few random things. Not being able to remove
the fsf-funding document isn't one of them.
So only one of your statements above is true.
By contrast, modifying the program to look exactly like another one, say
the source code to Windows solitaire, is already prohibited by copyright
independent of the license on hello.c, or whatever other program you
decide to start from.
Certainly, the Sun license goes about this in a different way -- using
its license to forbid you to go back to the original Sun RPC code and
distribute it -- but the result's exactly the same, and no more harmful
than copyright law in general.
If you want to get an opinion from Eben Moglen or similar on it, and pass
that on to Sun and the glibc folks, I'm happy to be convinced otherwise,
but as far as this goes, it's making an issue out of a triviality.
> Are you seriously suggesting that this is acceptable under the terms
> of the GPL? It's just another form of the invariant sections we've had
> so much fun with in the GFDL - data which cannot be detached or
> modified.
It's not this though -- it's the exact opposite. You can modify Sun RPC
to your heart's content; you just can't take out all the modifications
and end up with the single original source file Sun gave out.
> If not, please refine the statement "You need to be able to distribute
> the end product under the terms of the GPL, which you can; the
> original parts don't matter, since if you're distributing those,
> you're not bound by the GPL at all." appropriately.
Hrm? I can rephrase if you like?
If you take foo, that's BSD licensed, then anyone you give it too, no
matter what you might say about the GPL, can use it under the BSD license.
It's only when you hook foo up into some other code that you have the
copyright too (like Windows or glibc) that what you say about the license
is of any relevance.
> If this were true, then for all cases where multiple licenses applied
> to a given sequence of bytes, then you would have to obey _all_ of
> them.
I have no idea where you've pulled that claim from, but it's not correct.
If I take Capital-AJ 1.1 I can make Capital-AJ 1.1aj1 with whatever new
features I want. The only thing I can't do is the equivalent of:
cat Capital-AJ-1.0 | grep -v 'All rights reserved' |
dd of=Capital-AJ-1.1
and say "Oh, but it's not a copy of Capital-AJ-1.0, it's some changes
I made to Capital-AJ-1.1!"
> Under the GPL, I can take version 1.1, delete the first two stanzas,
> replace $word with aj and $worduc with AJ - simple modification to
> optimise the script.
And get sued for copyright infringement, no matter how vigorously you
wave your hands.
> I now have a copy of version 1.0 that is under the GPL.
No, you have a *copy of version 1.0*. Which concers the copyright of version
1.0, which you have no license to.
> At no point did I look at version 1.0 while performing this
> optimisation, so it is in no sense a derivative work of 1.0.
Considering version 1.1 has only ever been distributed along with 1.0,
that seems decidedly implausible. Considering the opitmisations aren't
valuable, and the features are (at least in comparison to the rest of
the script), you're almost certainly lying. Just as you would be if
you claimed the comments in SCO and Linux's mallocs were so similar due
solely to coincidence.
> If this did _not_ hold, then a corporation would be able to produce a
> GPLed program, and a set of modifications which you were not permitted
> to make - by implementing them and licensing them differently.
Except that they're not allowed to do that, because you can't change a
GPLed program and license it differently.
For BSD programs, Microsoft have already done such a thing: it's called
Windows, and includes a number of modifications to certain BSD code that
can't be made by anyone else.
Fortunately copyright law is mostly about expression, rather than ideas,
and it's usually easy to express the same idea differently, making this
a non-issue.
Cheers,
aj
--
Anthony Towns <aj@humbug.org.au> <http://azure.humbug.org.au/~aj/>
I don't speak for anyone save myself. GPG signed mail preferred.
``Is this some kind of psych test?
Am I getting paid for this?''
Attachment:
pgpxq559X5S6S.pgp
Description: PGP signature