Re: GPL traitor !

[David Kastrup]

Hyman Rosen <hyrosen@mail.com> writes:

> Alan Mackenzie wrote:
>> In (ii), the new code generator is an extension of GCC, using its
>> data structures, conventions etc.  This new code, if it is released,
>> must be released under the GPL.  Just how it's slotted into the
>> existing GCC is immaterial.
>
> You are wrong. Extensions fall under the GPL only if they incorporate
> GPLed code within them by inclusion, not by reference.

That's "if", not "only if".  The problem is whether the combined work is
more than a mere aggregation of its part so that the original
constituents can no longer be told apart well enough to be licensed
differently.

In that case, while the GPL does not actually extend its reach over the
other part, but it still demands that you need to license the _whole_
under the GPL or not at all.

The question is at which point the whole becomes a single entity.  Which
is not defined by the GPL, but by the courts and local jurisdiction,
drawing case law into account.

> That is, if the extension is written by taking existing code and
> modifying, it falls under the copyright of the existing work (and is a
> derivative work as well, so that the GCC copyright holders could not
> use it without the permission of its author). If the extension is
> written as a separate new work, it does not fall under the GPL, even
> if it makes extensive use of GPLed header files and of data structures
> that appear in the GPLed code. Attempting to use copyright to prevent
> interoperability is considered by the courts to be a serious breach,

A serious breach of what?

> and is not allowed.

There is no clear and consistent case law with regard to linking stuff.
When in doubt, you will try to make sure that decisions leaning either
way will not put you too much in harm's way.

-- 
David Kastrup