How To Break The GPL

W. Yip weng at
Sat Mar 4 18:57:03 UTC 2000

On Fri, 03 Mar 2000 15:56:06 -0500, John Cowan <jcowan at>
>"Forrest J. Cavalier III" wrote:
>> Says who?  If she distributed a derivative work of GPL'ed software,
>> then it must be GPL'ed.  The question is whether or not Alice has
>> a derivative work.
>In my first scenario, Alice made a derivative work but didn't distribute
>it.  She then distributed her own original work to Bob, who made another
>distributed work (identical to the first one) but didn't distribute it.
>Again I propose the analogy:  I mark up my copy of a certain copyrighted
>book, correcting certain errors in it.  I now mail you *only* the errata
>and corrigenda, which you incorporate into your copy.  Now we each have a 
>copy of the same derivative work, but nobody *distributed* the derivative work.

It seems to me this is the crux of the problem.

Alice may have

(i) infringed the copyright vested in Trent. 
(ii) breached the terms of the GPL

It seems to me that potential claims in the two cases vary.

In regards (i), Alice is unlikely to have infringed copyright because she
is a licensee under the terms of the GPL which expressly allow her to copy,
adapt, etc.

The problem tends to be in (ii). Alice MUST have agreed to the terms of the
GPL, or otherwise she would be infringing copyright as in (i). It follows
that in agreeing to the GPL, she has to fulfill its viral nature by
ensuring that any redistribution she makes, whether modified or otherwise,
is made under the GPL too. This is because the GPL covers distribution, and
that only, since it is a license that governs sublicenses.

John Cowan has pointed out the possibility that Alice can effectively cause
Bob to acquire Alice's derivation WITHOUT distribution. The issue here is
that since there is no distribution made, the GPL has not been breached.
Trent therefore can neither claim under (i) or (ii).

Subsequent contributions to this thread have suggested that the definition
of 'derivative work' may/may not encompass 'distribution'.

I submit that 'derivative work' as (i) a category of significantly changed
code is different from 'derivative work' as (ii) a manner of distribution.
It seems to me that (ii) is as yet unseen of, although John Cowans has
shown us how that may be possible. In regards linking, I think merely
linking or calling by function cannot be seen as distribution.

However, if we suppose Alice sends a file containing errata ONLY (no source
from the original) to Bob, can this be seen as 'distribution' of a
'derivative work'? 

I think the outcome of this program falls to how the courts interpret
'derivative work'. I think it really depends on how dissective the courts
want to analyse the program.


More information about the License-discuss mailing list