Opinions: linux and sound?

PhantomBeaker

Ok,

I was having a debate with a friend of mine - sound managers. I know KDE has its own sound server (arts) as does gnome and the like, but so far I've never successfully gotten mine working. As a result, it means that if I'm running my mp3 player, I won't get any console beeps etc.

And for once, I have to look towards windows as doing a vaguely better job.

Now the debate was that I figured the kernel should at least provide basic mixing capabilities. I.e. that multiple applications that access sound are at least mixed with equal weighting. Nothing more than a simple mixing equally from those sources.

His argument is that isn't the point of the kernel. Basically that handling that sort of thing isn't what the kernel is meant to do, and that that's what sound servers like arts are for. The fact that we have sound drivers in the kernel is enough. It's up to userspace applications to decide how to share it.

My counter to that is that the central point of the kernel is to share system resources etc in a reasonable manner.

But, I was wondering what ye all thought. Is it something that shouldn't be bothered with in the kernel. Is the fact that we just have sound drivers enough for us, and we should all arrange our applications to centre around one particular sound server? Or should we have some capability such that sound applications just use a certain device file and they don't have to worry about the existence of sound servers etc for mixing?

So, what do you think?

P.B.

Red Alert

many sound chipsets can do mixing directly in the DSP hardware such as the SB Live! and the Audigy. so you can open multiple connections to /dev/dsp AFAIK

Khannie

Not a kernel level responsibility IMO. Yes, the kernel's job is to share resources (and it does so), but the driver should be able to implement all the functionality that the card supports. My understanding is that the failing is at the driver level in this instance (had a few run ins with it in the past).

One of linux's biggest failings is the shoddy sound support IMO. Drivers have been around for god knows how long, but actually getting stuff to work is a nightmare. If I have xmms paused, I can't watch a video in mplayer. How ridiculous is that in a 10 year old, billion dollar invested OS?

Syth

The current situation with sound is appaling. I can't get 2 application to play sounds at the same time. If a basic sound mixer was included in the kernel, it would probably (hopefully) become dominant. The desktop experience would become much better.

tomk

TBH, I've no real idea what the kernel does, or how it does it, but my subjective response would be that it surely has more important things to do than sound mixing. I've often heard it said that "if it doesn't have to be done in the kernel, it should be done in userspace", or words to that effect, so would that not apply here?

My sound works very well for what I use it for. I haven't yet needed to, but I can watch a video and listen to an mp3 simultaneously - I don't have independent volume control (still working on it), but I can play/pause/stop each one without affecting the other. I don't use arts or esd, just alsa with dmix.

niallb

I've never had a good experience with the sound daemons,
but I'm happy enough with my sound cards.

I don't think the kernel should be managing the sound system at this level. Requests to play sound should be
treated in the context of a user session, as the user playing the sound should ideally be in the location where the speakers are. The kernel alone can't decide that.

I've been woken up in the past by a housemate accidentally trying to play sound through my speakers instead of his own just because he was logged in to my system.
The sound daemons can get around that, the kernel couldn't.

Like tomk, I just use the alsa drivers directly, and can play multiple sound applications.

NiallB

Big P

Linux does have built in mixing capabilities. The 2.6 series of kernels comes with alsa built in. You basically enable the driver for your card (which btw, is usually a hacked driver because the vendors do not release drivers for linux - nothing can be done about that). Once you have basic sound set up you create a ~/.asoundrc file. This uses alsa's built in dmixer application. You can find ~/.asoundrc files for your on the alsa homepage, or just google your card, linux and dmixer.

Of course you could also use other sound daemons like aRTs, jack etc. But dmixer provides the best audio experience imho. Plus its easy to setup.