[clug] The GPL and kernel modules

[Paul Wayper]

Chris Smart wrote:
> I had often thought that the idea of freedom was somewhat contradictory
> in terms of the GPL because I thought it was about ensuring that people
> have the freedom to do whatever they want with the software (i.e, there
> are no restrictions on what you can and can't do). But then I thought,
> I'm not really free if what I want to do is to incorporate it with
> non-free software. Shouldn't I be allowed to use the software as I
> desire without restriction?
>   
I was thinking about this, specifically in relation to the points you've 
raised on kororaa.org about including binary-only video card kernel 
modules with Kororaa.  And I was also thinking about how many people are 
saying "well, the arguments of the FSF, or Wasabi Systems, or whoever, 
are all very well, but they're just opinions - until it goes to court 
and a judge decides on it, then we don't really know what's going to 
happen."  I had two observations on this.

One is that what I _don't_ see is the FSF badgering nVidia and ATI, or 
taking them to court, over 'violating the GPL' by distributing 
binary-only drivers.  Nor (to my knowledge) are they badgering the Livna 
repository owners, for instance, for distributing binary-only RPM 
packages of the kernel modules for Fedora Core.  The FSF is quite happy 
to say to a distributor (Kororaa) that they should uphold the spirit of 
the GPL as they see it, but they're certainly not actually going to 
court to sort this issue out with the actual offenders.

Two is the more philosophical point that sorting these things out in 
court seems a stupid, backward way of doing things.  It's highly 
influenced by how much money each party has - i.e. what calibre of 
lawyers and legal research it can bring to the case.  It's highly 
influenced by the judge and the pitch of each team and what precedents 
they might have to hand then and there.  It's like saying that the only 
way to settle a border dispute is for two nations to go to war.  Surely 
there's a more positive, pro-active way to reach an agreement on the 
actual use of the license?

(I actually don't understand why nVidia and ATI and so forth haven't 
just come forward with the source code for their kernel modules anyway.  
It's not like the API for things like DirectX and OpenGL are 
particularly secret, and that's (to a greater or lesser degree) what 
their Windows drivers are implementing.  All their competitors will have 
disassembled their drivers anyway, so it's not like their raw hardware 
interface is either particularly interesting or secret.  And who bashes 
on raw hardware these days anyway?  Are there any software secrets that 
this source code is particularly protecting?  Surely the secrets are all 
in the hardware anyway...)

It seems to me that the fundamental issue with developing on GPLed code 
is it really is an "all or nothing" approach - either you share your 
code, or you don't use the GPLed code at all.  Which is not necessarily 
a bad thing, as the Linux kernel has proved: the fundamental fallacy 
that Wasabi Systems falls for is that giving away your code somehow 
makes you unable to sell it as well.  And despite Sarbanes-Oxley and the 
already strict terms of the GPL, the Sony rootkit has shown that 
companies are still copying GPL and other open source code knowingly - 
sabre-rattling from the FSF at small distributors is nothing if the big 
violators are still settling out of court and going on to violate the 
GPL elsewhere...

JM2CW,

Paul

P.S. By Torvalds, another 'short post' which turns out to be five 
paragraphs.  What the hell is wrong with me?