On Sat, 10 Nov 2007, Brian Buhrow wrote:
> 	Hello.  I'd love to see this.  I can think of times when when I could
> have used pcm audio data directly somewhere, but wanted to use the audio
> subsystem to get it.  It would be especially cool if you make it so that
> multiple pseudo audio devices can be attached.  Then you could do audio

This should already be possible.

> mixing through a hardware sound card by piping multiple data streams back
> to an open pipe to a hardware sound device.  (Probably trashes OS
> performance with all that copying in and out of the kernel, but we still
> don't have multiple voice support on most sound hardware, and the
> kent-audio project appears to be dead.)  This driver might obviate much of
> the need for that work.

That's very interesting. The only thing it doesn't give us is 
multiple-open support on /dev/sound, but this isn't a huge deal -- back 
when I was working with esa(4), I wrote an audiomux pseudo-device that I 
pointed to from /dev/{audio,sound}, and an open of this device would 
redirect to the first /dev/{audio,sound}N that was available.

Cheers,
Jared