Re: Pulseaudio

[Matteo Settenvini <matteo-ml member fsf org>]

On Tue, 2007-10-09 at 11:52, Gustavo J. A. M. Carneiro wrote:

> I don't care only about proprietary applications.  You think for example
> that Second Life Linux client (which is open source) will use Pulse
> Audio API directly?  It will take years before that happens.  I remember
> perfectly well how much time it took for applications to switch from OSS
> to ALSA, after Linux declared ALSA the official "blessed" Linux sound
> API.
> 

Pulseaudio does enable also to use applications still tied to OSS, btw.

> 
> It's good that there's an ALSA plugin to redirect sounds to the Pulse
> Audio daemon, although I must confess it doesn't sound entirely
> satisfactory.  Why be forced to use a userspace mixing program when
> hardware mixing would work equally well (or better) in most situations?
> Non-network audio should not need to be mixed by Pulse Audio on Linux,
> IMHO.

I'm not entirely sure, but I guess that software mixing is done only
when needed, using hw mixing where available. I may be wrong though.
Then PA could be fixed.

> 
>   Is there any good reason why Pulse Audio explicitly locks the audio
> device, unlike any other normal ALSA  client?  And no, making every app
> use Pulse Audio by force, just because you can, is not a good reason.

It may be, but not just "because you can". Having a unified interface
for GNOME would improve stability and uniformity. After all, why telling
people to use DBus and porting old applications to a new system, when
they could continue using ORBit? Why taking the pain to rework over old
apps to use GVFS and not continuing linking to GnomeVFS?

More, what if we want to port some GNOME app to a system that doesn't
have ALSA? dmix also has not so good latency support, and no sample
caching afaik. Also, it cannot control separately volume for different
applications (try adjusting your volume for a Totem movie to the maximum
and have Evolution ring the X11 bell or in a terminal and if you don't
become deaf you'll understand why this makes sense, but it's not just
that -- think about having foreground windows having higher volume and
bg ones lower, enabling for spacial sound effects).

Of course, switching to Pulseaudio is feasible just if it works 99% of
times, which it quite does for me (until now, I had only problems with
Wesnoth).

Writing a plugin or adapting pulseaudio to chew away that last 1% of
non-working apps isn't impossible, and we've time until 2.22 to fix it
quite well. After all, I had *much more* trouble with esound than not
with Pulseaudio, and esound is still shipped by default in a lot of
distros.

Then there is the point that Fedora and Ubuntu are pushing for PA.

Pulseaudio allows quite a lot of things that dmix doesn't. See also
http://0pointer.de/public/pulseaudio-presentation-lca2007.pdf.

Anyway, if you think Pulseaudio is bad, then GNOME libraries shouldn't
even try to link to libesd like they're doing as per now.

Cheers,
m.