For many years now, I’ve taken it as an article of faith that the GPL’s authors have been on firm ground when discussing the relationship of derived work, a legal concept, and of libraries, a software modularity concept.

Briefly, the FSF and many others interpret the GPL to mean that pulling a software library into a program means crossing the boundary of “derivedness”, so as to automatically make the program a “derived work” of the library … or vice versa. This is the reason that the FSF puts the GPL on much of its library-style software: to force applications to also be licensed under the GPL and thus become free software.

What’s new is a recent article in LinuxWorld (part 2, part 3). Its focus is on binary (or non-GPL) kernel modules for Linux, a somewhat murky area. The article argues that they are probably legal, because the notion of copyright infringement is different than most believe. The farther one reads the article, the clearer the implication that the popular layman’s inference of “static linkage => derived work => GPL infection” actually breaks at the first step.

If indeed courts are tending to consider as “derived works” in the copyright case law sense more and more tightly defined relationships rather than broadening them, focusing more on copying rather than dependency and especially “necessity”, then the whole GPL infection business is overblown. Rather than the FSF line (“doesn’t matter whether you link this GPL library dynamically or statically – you still have to GPL the application”), it may be the opposite (“doesn’t matter how you link … it still doesn’t force anything on the rest of the application.”)

There must be a simple retort to this interpretation. What is it?