GPL kernel clients

Bruce Perens bruce at
Fri May 28 04:37:55 UTC 1999

From: John Cowan <cowan at>
> So a Linux kernel on a VAX, where system and user code share the same
> address space, would bring every app running on it under the GPL?

And the i386 under Linux 2.2.x, too, but you're stretching the point beyond
absurdity. The kernel maps its client (the process) into the kernel's
address space, not the other way around. The kernel has a clearly-defined
IPC system that is used for all service requests, and it doesn't share
its program counter or privilege level with the client. Contrast this
to a shared library.

> I don't think shared address space can be the relevant parameter.

It's the currently-accepted test. Derivation is poorly defined in the GPL.
There are trivial ways of getting around the GPL by splitting a program into
a proprietary client and a GPL server. I've discussed this with Stallman.

> Note that the Linux kernel explicitly disclaims GPL tainting of
> applications that use it.

Right. It's a nice disambiguation, but since the implementation they're talking
about probably is a shared library, even though they consider it a "kernel",
it's more straightforward in this case to just use the LGPL.



More information about the License-discuss mailing list