Replace PulseAudio with OSS v4?

[Daniel Chen]

On Sat, Jun 20, 2009 at 1:47 AM, Danny
Piccirillo<danny.piccirillo at ubuntu.com> wrote:
> PulseAudio is actually a very bad choice in the long term due to horrible
> latency and lower sound quality, and that we should work to use OSS v4. It's
> a long read but seems to be worth it. What do others think about this?

Adding more layers inevitably results in increased latency if not done
correctly. PulseAudio's glitch-free mode addresses the interrupt-based
problem in a different fashion. Unfortunately, the state of Linux
drivers for common audio hardware in laptops is abysmal.

Yes, it's trivial to experience high latency using PulseAudio, but
that is not necessarily PulseAudio's fault. If you've seen any of my
presentations[0] on audio, you'll walk away seeing that Linux audio is
a complicated stack to troubleshoot and to improve incrementally.
Ubuntu has shipped with suboptimal configurations in the past, but
Jaunty was a fairly significant step forward (although many people
will dispute it because "sound is broken for me"). Karmic, by all
indications, will be better by virtue of more people spending cycles
fixing bugs in ALSA and PulseAudio. For instance, significant
buffering issues and audio anomalies have been identified and are
nearly resolved in the common case in Karmic, Rawhide, and elsewhere.
Closed-source software continues to be problematic.

Lower sound quality is a red herring. ALSA's default resampler has
known and quite audible limitations. The available resamplers in
PulseAudio demolish the "lower sound quality" FUD. Jaunty shipped a
configuration using a craptastic one in an attempt to balance CPU
usage with perceptive quality. Lessons learned: Karmic will ship with
a much better (but more CPU-intensive) resampler.

Now let's consider why replacing ALSA with OSSv4 in Ubuntu Karmic
would be a bad exercise:

1) No upstream mainline Linux support - Canonical and the Ubuntu
community would have to devote resources to supporting OSSv4 as
out-of-tree software, which is nontrivial for an area as significant
as the audio stack. The kernel team's lessons learned in supporting
such out-of-tree patches has indicated that no one would rather
continue down that road. To date, no one has stepped forward to
address the significant architectural concerns with merging OSSv4 into
mainline Linux.

2) Lack of feature parity - while some HDA codecs are marginally
better supported in OSSv4, that list continues to shrink. Creative
X-Fi support, USB, USB MIDI support, to name a few, are consistently
better supported in ALSA. Due to sheer momentum, that maintenance pace
does not hold for OSSv4.

That said, no one is opposed to seeing OSSv4 improve to the point
where it can be merged into mainline Linux. From the audio team's
perspective, it simply makes "support resources" sense for Ubuntu and
its supported remixes to carry support for ALSA and PulseAudio by
default.

I'd like to add that if someone wants to see OSSv4 support in Ubuntu,
that someone just needs to step up and work in the Ubuntu audio team.
I volunteer my spare cycles working on Ubuntu audio, so I see no
reason why a motivated and resourceful person cannot do similarly.

-Dan

[0] http://kernel.ubuntu.com/~dtchen/UDS-Barcelona/