Re: Recommendation for a CL data structures library

[Hyman Rosen]

On 3/25/2010 12:51 PM, David Kastrup wrote:
> If the program can't be compiled (and successfully prelinked) without
> inclusion of the corresponding library headers, it is somewhat strange
> to argue that the creation of the binaries is an act independent from
> the library, just because the _binaries_ of the library are loaded at a
> later point of time.

Copyright law is about copying protected works. When deciding on whether
a program file may be copied, the first question is do determine if the
file being copied contains any protected content. In a dynamically linked
executable, the only such content is what the compiler and linker have
placed into the executable in order to indicate that during execution the
program will invoke some specified functionality. But that cannot be
protected content because it lacks originality; there is generally only
one way to indicate that such-and-such a function will be called in such-
and-such a way, and any concept that can be expressed in only one way
lacks originality as defined by copyright law - the idea has merged with
the expression and the result is not copyrightable.

> You may be ferociously defending your own legal theories, but as long as
> nobody wants actually to rely on such a theory to a degree where he is
> willing to let himself be taken to court over it, that's academical.

In fact, such did happen with the distribution of a program which
dynamically linked to the GPL-covered readline library. To avoid
aggravation, the developer created a stub readline library, but it
cannot possibly be correct under copyright law for the rights to a
work to change by the creation of a separate work after the original
work has been created!