[linux-elitists] ATI Radeon support

Ben Finney ben@benfinney.id.au
Mon Dec 13 18:36:22 PST 2004


On 13-Dec-2004, Jason Spence wrote:
> On Tue, Dec 14, 2004 at 12:09:04PM +1100, Ben Finney wrote: 
> > I'd settle for them releasing full 3D specs, if they don't want to
> > release the software.  I don't care who writes the drivers, so long
> > as they're free software and give full access to the features of the
> > hardware.
> 
> I agree with your philisophical position.  However, I believe that
> getting full access to the hardware is a losing proposition, because
> it's very difficult to do anything with raw access to the hardware
> that isn't highly specific to your particular card.

There's a distinction in there that I was obviously not careful enough
with.  I originally wrote "full access to the hardware", just as you
did; but I realised that's not necessary.  I then amended it to "full
access to the features of the hardware", which is quite different.

It's not necessary to get access *directly* to the hardware; just to be
able to *do* anything that a user of the card can do with the non-free
drivers.  If that means an API to the card, fine; I'm not asking for
access to the processor registers or anything beyond what's needed to
*use* the thing.

Sorry for not making that distinction explicit when I tripped over it
myself.

> nVidia, at least, provides an internal abstraction layer in their
> drivers which hides all the changes they make from card to card.

Sounds like a great approach.  Publish the full API to that abstraction
layer, under terms usable by free software developers, and we're done.

> I think it's best that the hardware companies do this kind of thing.
> Creating the driver yourself from the specs would involve a massive
> amount of work that could take years to pay off, by which time the
> interfaces you're programming against will probably have become
> obsolete.

If they stick to a known API for drivers, it's a great thing.  It levels
the field, so that any kernel -- even ones that the manufacturer has
never heard of or has no interest in directly supporting -- can have
equal access to the hardware.

> For full OpenGL compatibility, for example, you'd have to write a
> shader compiler and fixed pipeline state optimizer, both of which are
> very complex components.  Also, the community would be trying to outdo
> some of the best 3D driver people in the world!

The community is open, they can join any time.  There's no need for the
best 3D driver people not to be *in* the community.

> I also think that it would be better if there was some kind of feature
> that the community could provide in their driver set that the hardware
> companies couldn't;

Again, I don't see why it has to be us versus them.

> > Care to elaborate?  What is it about one party releasing free software
> > drivers, that the other would sue for?  Do they each have NDAs with the
> > other, or something?
> 
> Maybe there's some shared intellectual property between them, but
> they're probably worried about being sued by one of the other parties
> they license IP from.  Consider:
> 
>  - I license some IP (say, a corner-case culling algorithm) to nVidia,
>    which they incorporate into their driver.

You've identified the mistake.  It's their own choices keeping them from
sharing.  I've no idea why nVidia insiders keep crying about how they
have no choice in the matter.

> There goes my revenue stream!

I have no interest in protecting the interests of those who want to
divide software developers and users, and prevent them from sharing.

I take your point that this may be what they're afraid of; however,
they're the ones who chose to be in this position, and I can't see how
anyone else except them can do anything about it.  Hence, the continued
beating on their door to *fix* the problem.

-- 
 \      "I hope some animal never bores a hole in my head and lays its |
  `\   eggs in my brain, because later you might think you're having a |
_o__)          good idea but it's just eggs hatching."  -- Jack Handey |
Ben Finney <ben@benfinney.id.au>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: Digital signature
Url : http://allium.zgp.org/pipermail/linux-elitists/attachments/20041214/a11ee887/attachment.pgp 


More information about the linux-elitists mailing list