Horst H. von Brand wrote:
> They claim that it doesn't matter if the components are
> distributed separately or not. For example, you can't modify a GPL'd
> component so that it needs a library under different terms even if the
> parts are never distributed together.
Sure you can. In the heyday of GNU software it was /only/ run on propietary
systems, and all that software was routinely modified with the express
purpose of running on newer/changed versions of propietary systems (which
were the only ones available, remember).
There's a special exception for the case of standard system software
that doesn't apply in other situations - or if they are distributed
> Whether it is a derivative or 'work as a whole' depends
> relationship of the parts, not temporary physical proximity.
Please state which laws or court decisions lead you to this conclusion, and
exactly what "relationship" has to be involved here.
I'm not aware of any court ruling. It is the conclusion I drew from the
FSF legal position regarding RIPEM that lead to the development of the
generally useless fgmp library. It was long enough ago that google
doesn't find many good references, but this one at least describes it: