Re: [Dri-devel] Re: Progress of mesa-3.5 tree? Update to Mesa-3.6/4.0

[Mark A. <ma...@em...>]

Will Newton wrote:
> 
> On Friday 14 Sep 2001 9:25 pm, you wrote:
> 
> >  Is open source absolutely essential? Personally, I would rather have
> > binary-only drivers written by the likes of Brian, Gareth, Keith, et.
> > al. than binary-only drivers written by some faceless unknown.
> 
> Binary only drivers, ups and downs.
> 
> Ups:
> 
> The driver exists.
> 
> Downs:
> 
> It may not work on your architecture, with your kernel.
> You can't debug it, or any other part of your kernel.
> You can't audit it for security holes (i.e. don't put an nVidia card in a web
> server)
> It sets a precedent, and a bad one.

 There's no good reason why you couldn't put an NVIDIA card in a sever
and use the open source 2D driver. You wouldn't want to enable DRI on
you server, either, because of lockup issues. 

 You may feel that the NVIDIA driver sets a bad precedent, but it also
sets a high standard. Their drivers are high performance, complete, and
stable. They have several people with histories in open source 3D
efforts, people who care about 3D on Linux. You trusted these people
with your drivers when they worked on Utah and DRI, why shouldn't you
trust them when they work for NVIDIA?

 If binary-only is the only way to generate enough revenue to support 3D
driver development for Linux, it should at least be considered.

> >  Would an open/closed hybrid be feasible? A bare-bones, rasterization
> > only base implementation could be completely open source, included in
> > distributions, etc. A fully-featured driver w/ interesting extensions,
> > T&L, etc. would be a binary-only, licensed product. Something similar to
> > the way OSS works.
> 
> OSS works, but only just.
> How many people do you know who bought OSS drivers?
> How many people do you know who use ALSA instead?

 OSS supports a whole lot more sound cards than ALSA, especially
professional cards. ALSA is a viable alternative these days, but it has
only become stable enough for widespread usage in the last year or so.
Without OSS, Linux would have been without sound for a long, long time. 

 I have purchased the OSS drivers (for cards that weren't supported by
OSS/Free or ALSA), and have been very pleased with the ease of
installation and support. They have a model which allows them to make
enough money to support themselves and write new drivers for cards that
don't have public specs. 

 The important thing is to have drivers available from a source you have
some confidence in. If Linus wrote a binary-only kernel module that
provided some functionality you require, would you have faith in it?
Binary only support is usually favorable to no support at all. 

> >  What other options are there? Is there a way to make this work? It's
> > worth a bit of brainstorming and thought about whether some model can be
> > put in place to support 3D driver development.
> 
> Step one: Get specs.
> 
> It's that simple. If you have enough people with ability and the specs, you
> can get somewhere. No specs? Forget it.

 No, it's not that simple. How long have the G400 specs been available?
I still have lockups and general flakeyness with DRI/G400. Stable
drivers don't just magically appear when specs arrive. You need talent,
and you need time. Even with the amount of talent working full time on
the DRI project, things have been moving slowly. It's a big, difficult
job that needs a lot of attention to do correctly. 

 All the open source cheerleading the in world isn't going to make
drivers appear from thin air. 

> >  3D is becoming increasingly important for general PC use. In the past,
> > the domain of 3D has been primarily for DCC and games. In the future
> > (and not that far in the future), 3D will be pervasive throughout the
> > GUI. Without good 3D support, Linux is dead.
> 
> Never seen a 3D office suite/web server/ etc.

 You can, and you will. Microsoft Office has 3D support. Star Office has
OpenGL support for charts and graphs. 3D support in office suites is
only going to increase because it allows you to create flashier
presentations and more complex plots and graphs.

 3D acceleration will also have an increasing presence on the web. It's
pretty clear that VRML isn't going to make much of an impact, but Flash
uses OpenGL now. There are a lot of applications for 3D for web
traversal, etc. There's going to be a lot of content you won't be able
to access if you don't have 3D. As for servers, what about vis-servers
which render large-scale visualizations on the server then pass the
final image to the clients? You may not come into contact w/ these
machines, but they have a lot of applications in scientific computing,
medical imaging, oil/gas exploration, etc.

 Next generation GUI interfaces will leverage 3D as well. 

 Just because you don't see a lot of it *today* doesn't mean that it
won't be important tomorrow. It will.

-Mark
------------------------------------------------------------
Mark B. Allan                   NASA Ames Research Center
QSS Group, Inc.                 Neuro-Engineering Lab
650 - 604 - 0537 (office)       Mail Stop 269-2
650 - 604 - 0461 (lab)          Moffett Field, CA 94035
650 - 604 - 3594 (fax)          ma...@em...
http://ic.arc.nasa.gov/ic/ne.html
------------------------------------------------------------