[ILUG] treatise on non-free/binary only drivers

[John Gay]

On Thu 02 Sep 2004 21:21, ilug-request at linux.ie wrote:
> It's subscriber only content on lwn for the next week. I'm not going to
> reproduce it here for various reasons, but I'll copy and paste little
> bits of it.
>
> Quotes:
> The first of these is that the copyright status of many of these modules
> is ambiguous at best.
Which could explain the reason for the binary modules, as I'll explain below.

> Binary modules are, by their nature, platform-specific.
> When binary modules have bugs, there is no way to even track them down,
> much less fix them.
Fixing bugs in the binary module is the task of the supplier.

> Closed-source modules break when the system is upgraded.
And the supplier needs to debug and fix it when it does.

> Binary-only modules lack transparency;
That's the purpose of binary-only modules, to protect IP.
>
> Rephrasing:
> Binary only modules hold back development of the kernel. When a new
> feature comes along, if it breaks binary only drivers, users complain.
The users need to complain to the supplier of the binary module, as they do 
when any other part of their system breaks. Post the bug reports to the 
maintainer of that portion of software.
>
> Their conclusion:
> It is thus in the interest of all users to discourage proprietary
> modules. It is not a question of irrational allergies to end-user
> license agreements or free software fundamentalism; it is, instead, a
> matter of creating the most stable and capable kernel possible.
Stable, yes, capable, no.

This is the point of view from the kernel developers. But imagine the problem 
from the binary-only module suppliers POV:

The Linux Kernel is an extremely fast moving target to try to follow with 
little protection from major ABI incompatibility issues.

Not everyone actually owns the IP they use. This was high-lighted with the 
Sorensen codec. Apple had bought a licence for part of the code from another 
supplier. neither one could release the specs without the others permission, 
and each was waiting for the other to move first.

Second example: When Sun first bought StarOffice, large portions of it were 
the IP of other companies. This code had to be stripped from the first 
release of OpenOffice, The Spell checker, thesaurus, Adabas Database and 
others. The open source community worked very hard to replace this 
functionality before OpenOffice could even build properly.

So it's entirely possible that companies like Nvidia, ATI and other suppliers 
of binary-only modules don't have the ability to release the specs or source 
code even if they wanted to.

In fairness to Nvidia, they have gone to great lengths to keep their drivers 
up-to-date with the kernel. Also, they have created a unified driver so that 
improvements in drivers are available for all platforms rather than developed 
on one and back-ported to the others. I realise binary-only drivers are not 
the best idea, but sometimes they are the only option. 

Would you prefer if Nvidia stripped out a large chunk of functionality from 
their video cards to be able to offer source code for a partially functional 
card?

Sometimes it helps to put yourself in the other guys shoes for a while.

Cheers,

	John Gay