[NCLUG] what constitutes a derivative work?

[Jake Edge]

Regardless of the merits, or lack thereof, of the SCO case in SCO vs. IBM, 
I think there are some interesting questions regarding derivative works.
I have struggled with this issue several times regarding GPLed software.

Based on the theory that seems to underlie the SCO case, if I have seen 
code that they have copyright on and then go on to develop code that works
with their copyrighted code (or has substantially similar functionality),
it is a derived work and therefore theirs, even if I don't use a single
line of their code.

Now, most folks on this list have seen the Linux kernel code and if 1 or
more of us was to go off and write a kernel that was not GPLed, but was
not using any of the Linux code, is it a derivative work?  Does the 
supposedly viral nature of the GPL come into play (given SCO's claims, I
don't really see how the GPL is any more viral than other copyrighted
code, but I digress)?  If I develop a device driver for Linux, is it a
derivative work (Linus says no, but he doesn't own all the copyrights)
and what about a kernel module that isn't a driver but uses kernel services
through the established interfaces (even more of a gray area, but I believe
Linus still believes it is not a derivative work)?

If this is the case, how can one ever decontaminate oneself?

jake

-- 
Jake Edge - Edge Software Consulting, Inc. - jake at edge2.net