[clug] The GPL and kernel modules

Cameron Patrick cameron at patrick.wattle.id.au
Wed Jun 21 04:24:47 GMT 2006

Conrad Canterford wrote:

> The point is - what you, or Linus, or Richard Stallman (sp?) interpret the
> GPL to mean doesn't mean anything. What it means can ONLY be
> (conclusively) decided by a court.

*nod*  As far as I know, all of the lawyers (and legally-aware
individuals) who have considered the GPL and written about it interpret
it similarly in regard.  As far as I know, that's the closest we've got
to an actual court decision here.

> This is a nice theory. It might even work. But I certainly wouldn't be
> betting any money on it. One of the biggest problems large scale GPL
> projects are going to face is that it is a collaborative effort. All you
> need is for one developer out of a thousand to say "Oh, but *I* don't mind
> my code being used by non-GPL'd programs" and suddenly you have a horrible
> horrible mess to sort out.

Contrariwise, all you need for the spam to hit the fan is just one
developer saying "you're following the terms of my licence when you're
creating derivative works of my software" and the wherewithal to take it
to court.  Paul Wayper's comments about settling these disputes
_without_ resorting to the legal system are spot on here, I think.

> Personally, I don't see a problem with binary kernel modules provided some
> basic rules are followed. They are no more a "derivative work" than any
> piece of commercial software that runs on Linux, MacOS or Windoze.

The process of loading a kernel module is exactly the same as the
process involved in linking a library to an application, which is
exactly the same as the process involved in linking together two source
files in the _same_ application to form a single executable.  That's
almost always going to result in a derivative work.

Userland software calling the kernel, on the other hand, does so using
specific machine instructions designed for that purpose (e.g. SYSCALL on
x86) and uses precisely specified interfaces which are more or less
isomorphic to those found in other Unix systems.  That's pretty hard to
conceive of as being a derivative work of Linux specifically.


More information about the linux mailing list