Is inherited class a derivative work?

[Michael Beck]

We have a discussion going on internally about inheritance. Some people believe
that when you subclass/inherit/derive a new class, you are creating a
"derivative work" in the copyright sense, especially when you override existing
methods. Others believe that "inheritance is delegation of implementation and
aggregation of interface", and therefore it is NOT a derivative work.
Interestingly enough, at:
http://www.nusphere.com/products/library/gpl_0401openmag.pdf
Laurence Rosen states that:
--------------------------
A derivative work is not created by merely touching any more than one catches
AIDS by merely hugging. In copyright law terms, one must create "an original
work of authorship." An author must consciously recast, transform, or adapt the
GPL-licensed software (all of which are forms of "modification") before the GPL
applies to the new software.
--------------------------
which would suggest that a derived class, as an "adapted" class, is a
"derivative work".
Also commonly calling the new class a "derived class" would indicate that it is
a "derivative work".
The issue is related to following: If we release a software under OpenSource
license, we want that any modification to the software and any derivative work
go back to the community. The scary scenario is that somebody will inherit a
class, make some modifications to it, and then claim that since it's not a
"derivative work", s/he doesn't have to publish it, and the new class is his own
proprietary code.
Who is right in copyright terms?
Michael

--
license-discuss archive is at http://crynwr.com/cgi-bin/ezmlm-cgi?3