From alsa-devel-owner@alsa.jcu.cz  Wed Jan 13 22:22:35 1999
Received: from cgmd228195.goldmine.nl (IDENT:rutger@cgmd228195.goldmine.nl [195.75.228.195])
	by marvin.jcu.cz (8.9.1a/8.9.1) with ESMTP id WAA32293
	for <alsa-devel@alsa.jcu.cz>; Wed, 13 Jan 1999 22:22:20 +0100
Received: (from rutger@localhost)
	by cgmd228195.goldmine.nl (8.8.7/8.8.7) id WAA02616;
	Wed, 13 Jan 1999 22:22:18 +0100
Message-ID: <19990113222217.B2545@spin.m38c.nl>
Date: Wed, 13 Jan 1999 22:22:17 +0100
From: rutger@spin.m38c.nl
To: alsa-devel@alsa.jcu.cz
Subject: Re: as if to prove my point ...
References: <199901130726.CAA22126@op.net> <Pine.LNX.3.96.990113131127.15135C-100000@entry.jcu.cz>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <Pine.LNX.3.96.990113131127.15135C-100000@entry.jcu.cz>; from Jaroslav Kysela on Wed, Jan 13, 1999 at 01:34:13PM +0100
Reply-To: alsa-devel@alsa.jcu.cz
Sender: alsa-devel-owner@alsa.jcu.cz
Precedence: list

[snip]
> 
> Jacob Crawley contacted me directly and we are looking for way howto
> support ALSA driver development for SB Live which should be acceptable
> by both sides. Looks that Creative can release some code under free
> license (it willn't be probably GNU GPL, but something like Netscape
> agreement for Mozilla). Main problem should be that Creative is thinking
> about release only some library for SB Live - this should cause unwanted
> trouble with different kernel versions :-((
> 
> 							Jaroslav

On news://news.3dfx.com/glide.linux something similar was going on: a
company wanted to create driver but didn't want to release source
code, and then the next thing was suggested which seems as a workable
solution:

From: Jason McMullan <jmcc@ontv.com>
Subject: Re: Real Linux 3D cards
Date: Friday, January 08, 1999 4:24 PM

Luke B. Bishop <nanosoft@nanosoft.hypermart.net> wrote with confidence:
> Anyways, I just got a response from the S3 developer relations (still
> nothing from 3Dfx...).  They just sent me their standard NDA's.  It
> looks like there is a possibility of a Savage3D driver for linux. 
> Anybody want to comment on this?  What form it should take, etc?

  How about a .o that provides all the accelerator functions, 
has a documented .h file for said `bare minium hardware API' 
and has these properties:

	a) Compiled as PIC so it can be used in a shared lib
	b) Make no kernel or C library calls. 
	c) If it _must_ make an external call 
	   (ie mmap()/phys_to_virt()/etc.) put them in a
	   different namespace (ie shim_mmap()/shim_phys_to_virt())
	   and place their implementation in a different (and
	   distributable) .c file

  With these properties, you can make a SINGLE binary .o file
that can:

	a) Be `wrapped' with the appropriate kernel
	   functions and be used by FBcon
	b) Be `wrapped' with KGI and used by GGI's kernel side.
	c) Be linked into a shared library (ie MesaGL)
	d) Be directly linked into a program (ie demos)

  I have used this technique before with much success, and many
of the GGI KGI kernel drivers are tested in user-land before
going into the kernel with techniques similar to this.
	   

-- 
Jason McMullan - Linux - GGI - http://pepsi.visus.com/~jmcc
  I can't wait until the point where free Linux CD's or other paraphenalia is 
included in cereal boxes.  That truly will be the cherry ontop of our madness!
  -- Mike A. Harris <mharris@ican.net> on linux-kernel@vger.rutgers.edu

-- 
Rutger Nijlunsing, rutger @ null.net ----------------------------- Linux! --
Don't BiCapitalize without extremely good reason: it messes up the natural
human-eyeball search order -- Your Friendly Neighborhood Archive Maintainers
+31-40 ----------------------------------------------------------- ^X^S^X^Cs

