Alessandro Rubini wrote:
Actually, I've never heard the FSF claim that the _source_code_ of a
program using a (black-box) library is derived from the library. What
it claims is that the executable is derived from both,
Maybe there is some confusion here between "derived" in everyday
language and "derivative work" in copyright law?
It is, I think, as certain as any law can be that copyright (at least in
the US) requires an original work of authorship, which requires an
author and a creative input. I think it is fairly certain worldwide that
a derivative work requires the same.
A compiler can only perform a transformation from source to object form
programmed into it by its creators; it is neither an author nor capable
of creativity; it can this not produce an original work of authorship or
thus a derivative work.
If A is derivative work of B, then the compiled form A' is probably too.
If A is not a derivative work of B, then A' is not either.
I'd be surprised to see the FSF argue something other than the above.
Then again,
http://www.gnu.org/licenses/gpl-faq.html#DevelopChangesUnderNDA seems to.
and the authors
of either part have their say in choosing distribution terms.
That is quite true. After all, you can require any condition (within the
law) you want to even allow people to copy the software; however, such
conditions may (of course) not be free.