Replace PulseAudio with OSS v4?

[Davyd McColl]

Personally, I would welcome just about anything which would help us to lose
PulseAudio. Or magically transform PulseAudio into something which doesn't
suck. Either way would be fine. Allow me to elaborate (or skip the rest of
this post if you don't care):

I've had an SB Live for ages. One of the most redeeming features of this
card is hardware mixing. Meaning that I didn't care about OSS lockups or
ALSA's dmix. Things "just worked". Most users like it that way. Recently,
when loading Win7 to be able to play some windows-only games, I've found
that windows hasn't had proper SBLive support since, well, XP. XP picks up
the card on my system but doesn't output sound to it. Win7 seems to think
it's a relic from a distant age and refuses to work with it. Creative,
apparently, don't care. So the card that I've used for years because of how
it rocks under Linux had to go -- I want a system I can just reboot to play
my games (that is all windows is good for, imo).

I tried using PA's mixing and multiple output to use USB headphones and the
onboard Realtek HDA audio. Worked for a while but often left PA locked up. I
would have to kill and restart. My nett conclusion is that PA doesn't do
well with multiple soundcards, despite the advertisements.

So now I use the onboard sound exclusively. PA behaves (mostly) for me, but
the sound is a little latent -- and I'm not a person who creates music or
anything like that. I can deal with the minor latency because it doesn't
really affect me. Someone who mixes digital music on the other hand (and I
have a friend who does) can't use PA.

Now, when mixing wasn't an issue (ie when I had my SB Live), OSS was all I
needed. Apps which wanted ALSA would also work because the kernel supplied
the API. But ALSA didn't give me anything I needed. Then again, neither
would have done the multi-card output seamlessly. I guess I have to agree
with the general consensus that sound is not Linux's stronger suit. I guess
it comes back to my initial comment: I would welcome (and I'm sure other
users would agree) any subsystem which:

1) Worked (all the time, without random lockup)
2) Wasn't latent
3) Wasn't a mission to set up
4) Just handled mixing -- it's not something the average user thinks about
when Redmond has never really made it an issue -- multiple win32 sound apps
have just been able to work simultaneously since, well, almost forever.
5) Could handle multiple soundcards easily -- those USB headphones might
still come in handy instead of the extension cables from my onboard sound
(my keyboard has a USB hub on it -- it was well convenient).

Personally, I have yet to see that list met by any system. OSSv4, from the
posted article, looks like it handles the average user's requirements quite
well. I guess it's up to whether it's worth patching into the Linux kernel
for *buntu distros or if the kernel devs want to include it. On the other
hand, I have, in the past, after much frustration, managed to get ALSA's
dmix to work -- oddly enough, some distros actually have tools to make it
work for you. I haven't seen something like that on *buntu (though I have to
admit that I didn't look *too* hard because those were the days of the SB
Live).

It would indeed be a great step forward to have sound work under Linux in
the same manner that windows users are accustomed to: it just does (barring
stupid sound card providers who drop driver support, of course).
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/ubuntu-devel-discuss/attachments/20090620/72ac755b/attachment.html>