[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Netatalk and SSL



Ben Pfaff <blp@cs.stanford.edu> writes:
> Russ Allbery <rra@debian.org> writes:

>> Because the Linux kernel adds an additional clause, in the form of a
>> statement of the author's interpretation of the license, saying that
>> such modules are okay.

> Are you saying that the NVIDIA driver for Linux is a user program, not a
> kernel module?  (I do not know for sure because I have never had cause
> to download or install it.)  Here is the clarification included in the
> COPYING file distributed with the Linux kernel.  It does not talk about
> kernel modules at all, only about system calls made by user programs.

>       NOTE! This copyright does *not* cover user programs that use kernel
>     services by normal system calls - this is merely considered normal use
>     of the kernel, and does *not* fall under the heading of "derived work".

Oh, hm.  Good point.

I'm so used to Linus talking about the reasons behind his exception for
modules that I had forgotten that they weren't mentioned in his exception.
Sorry about the misleading comment.

One can extend the above argument by way of observing that Linux has a
clear interface layer that could be treated the same as the system call
layer, but Linus doesn't explicitly say that, and the same could be said
of just about any shared library.

Probably the more persuasive argument, which *also* applies to binaries,
is that Debian is not doing the mixing of the NVidia kernel module and the
kernel; the user is doing this when running insmod.  The user can create
works that are not redistributable without violating the GPL; the GPL does
not limit what you do with the pile of code that you have on hand as long
as you don't give it to someone else.  However, I must say that it's hard
to see the distinction between distributing a separate kernel binary and a
kernel module and claiming the user is doing the combining and
distributing a separate compiled binary and a library and claiming the
same thing.

(A particularly good edge case, which has some serious practical effects
in Debian, is whether one needs to care, as an application author, about
the licenses on system nsswitch and PAM modules that are dyanmically
loaded into the same executable based on a configuration controlled by the
user.)

-- 
Russ Allbery (rra@debian.org)               <http://www.eyrie.org/~eagle/>



Reply to: