X and Eye Candy

[D. Hugh Redelmeier]

| From: Evan Leibovitch <evan-ieNeDk6JonTYtjvyW6yDsg at public.gmane.org>

| If the binary drivers are freely copyable then they will be around
| forever, able to work with X servers that don't change the interface.

Those are big "if"s.  In the case of the kernel, Linus has expressly
reserved the right to change the kernel/driver interface.  For video
cards, some of the support is in the kernel (DRI).

Even X is likely to change (after shaking off a decade of
hibernation).

So no, a binary video driver is not good enough.

| However, you offer a hint at something here...
| >Old ATI cards have open drivers that can do 3d.  This is not the case for newer ATI cards nor for nVidia cards.
| >
| ATI makes no new money off old cards, neither does nVidia. The
| proprietary edge that they must keep on current cards has a finite life.
| One would hope that as the next generation of cards is available, source
| for the no-longer-sold ones will be opened. This tactic has a reasonable
| chance of success and is a reasonable middle ground.

I have no knowledge of why the card makers withhold specs.  Depending
on the reason that is operative, they may release specs for old cards.

Plausible reasons off the top of my head:

1) patent infringement suits might arise if opening the source makes
   it easier for patent holders to see that the card might be using
   techniques covered by patents.  This sounds farfetched to me -- as
   an idealist, I expect that the companies pay for all patents they
   use already.  On the other hand, I've seen this put forward more
   than once.

2) trade secrets
2a) trade secrets of the card maker that they wish to keep secret
2b) trade secrets of other companies that the card maker is bound by
    contract to keep secret

3) the card is so ugly and buggy that the card company wants nobody to
   see it.

4) the card "works for me" for the drivers that the company wrote.  If
   the company releases a spec, an independently written driver will
   exercise the card in new ways, exposing bugs to be analyzed and
   dealt with, ones that did not matter if the specs were not released

5) conspiracy with Microsoft to keep FLOSS disadvantaged.  This would
   benefit Microsoft and might consider it a favour.  The card
   companies are very heavily dependent on Microsoft so Microsoft has
   tremendous leverage.

   + MS designs Direct X and can advantage or disadvantage a card
     severely in this process

   + MS makes X Boxes and buys a lot of video chips for them

   + MS makes Windows XP Media Center Edition and thus can advantage
     or disadvantage TV tuners.  ATI's All In Wonder was heavily
     advantaged early in this game.

   + MS distributes drivers through Windows Update.  This might
     matter.

   + MS certifies drivers and cards for WinXP

6) FLOSS does not appear to be critical to the success of a card so
   management deploys its resources other places.

7) Card makers want old cards to die so that they can sell new ones.
   Anything that extends the life may cannibalize sales of new cards.

[additional reasons go here]

Few of these reasons go away when the card becomes obsolete.  Perhaps 2a.  
But I don't even believe 2a is a rational reason in the first place -- 
other card makers are surely capable of reverse engineering without driver 
source.

I am currently very annoyed at ATI for not supporting the x1000 family
on Linux, even with their proprietary driver (even though their web
site implies that it does).  See
  http://www.driverheaven.net/~pete/article5.htm
My contribution to the discussion of this article is
  http://www.driverheaven.net/showpost.php?s=88050b62bd52cc27a0f32b66475087b9&p=815010&postcount=51
--
The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml