On 10-Jan-04, at 8:28 PM, Perry E. Metzger wrote:
   
   > It would be nice if we built a kernel mixer facility so that the
   > device COULD be opened more than once. Frequently one wants to listen
   > to to an mp3 and still get console feeps and the occasional sound from
   > a chat program. That would give us a more general way to fix this...
   
   There are a few hardware drivers that support this functionality today 
   (esa(4), eap(4), maybe emuxki(4) as well (I know the hooks are there)), 
   but you're right -- a generic in-kernel software audio mixer would be a 
   great feature to have.
   
   With esa(4), I run artsd on audio3, mp3s on audio2, and that leaves me 
   two audio devices free for whatever else I'm doing at the time.


oh, i agree that for hardware that supports *hardware mixing* we
should make it so /dev/audio can be opened multiple times.  that's
just a really really really really big change to make to the system
(changes device close semantics of course.)  this sort of system
_completely_ blows away any hack with artsd or so...but these tools
are still required for single-signal devices...


however, IMO, there should not be any code to mix audio in *software*
in the kernel...



.mrg.