[Swfdec] Resurrect OSS support for non-ALSA operating systems

[Chris Wareham]

Benjamin Otte said on 17/10/08 11:59:
> Hey,
> 
> Here's the necessary knowledge about audio backend choice:
> 1) I don't want to maintain multiple backends, because identifying
> bugs in backends I don't use is rather hard. So I try to find the best
> one possible for me (I run Ubuntu) and make it work. That is currently
> libalsa.
> 2) The long-term goal for Swfdec is to switch to Pulse, as it seems to
> work anywhere. However, considering that Pulse has no good API to
> write to currently[1], I've not yet looked at making the switch.

Yikes. Pulse and ALSA both seem to be inferior compared to OSS. It's a
shame that OSS licensing lead Linux to adopt a completely new audio
subsystem rather than do what the BSD's did, which was to continue with
the last open source OSS version and improve it.

> 3) The current Pulse backend is written by Eric Anholt, and he uses it
> on FreeBSD. He's not very actively hacking on it, so I guess it might
> contain various bugs.
> 
> That said, I can think of 3 ways forward:
> 1) Use Swfdec's Pulse output and make it work.

Possible problems with Pulse notwithstanding, this appeals to me, and I
understand perfectly your desire to support only one audio backend.

> 2) Make ALSA work on BSD (probably without direct output through OSS
> or Pulse or whatever). Not sure how hard that is, but it'd port you
> lots of other apps, too.

Creating a BSD ALSA compatibility layer would be great - as you point
out it would open the way for applications like Rosegarden to work - but
it's a considerable undertaking. A while ago I wrote a NetBSD
compatibility layer for raw MIDI, which is a very minor part of the ALSA
API, but never got it working satisfactorily.

> 3) Revive the OSS backend.
> 

If Pulse wasn't an option, then supporting both ALSA and OSS would
appeal to me. As things stand though, I'll carry on trying to get swfdec
working with Pulse on NetBSD.

> I personally would prefer option 1 or 2. 1, because it would prepare
> our Pulse backend for the switch better. And 2, because it would solve
> lots of BSD audio issues and make life easier for you guys. Option 3
> is not that interesting for me as I think supporting OSS is like
> flogging a dead horse.
> 
> Cheers,
> Benjamin
> 
> 1:http://linuxhaters.blogspot.com/2008/10/pulse-my-audio.html
> 

Regards,

Chris