[Wine] Re: Pulseaudio

oiaohm wineforum-user at winehq.org
Thu Oct 1 08:26:52 CDT 2009

Ossp is not a kernel patch.  Ossp is userspace and depends on a kernel feature cuse(char devices in userspace)  Cuse appears in 2.6.32 kernels end of year kernel release.  Ubuntu 10.10 release cuse will be a default feature no kernel building required.  Same with every distribution using 2.6.32 and later kernels.  No point wasting effort for something that is only going to be useful for a few months.

ossp is very much the windows user space sound server model.  Kernel supported interface that all applications talk to.   No direct api to sound server no bipass.

Windows has a userspace sound server Oli Warner a kernel backed userspace sound server.  So applications cannot bypass it and there only can be 1 at any 1 time.  I should have been more correct all sound servers on Unix/Linux/BSD have had the defect of being bipassable by design.

There is a correct way to build a perfectly stable sound server and a incorrect way.  Pulseaudio is the incorrect way.   Since it is the incorrect way at some point it will have to be replaced.  As long as it does not take 20 years of glitches for developers to get around to doing it.

Basically if you stay with pulseaudio you are always going to have glitches from time to time until they address the interface issue. 

For wine to support Pulseaudio directly you have to prove to us that its not a defective design that we will have to do a new driver in future.  Pulseaudio poor ALSA and OSS support has caused issues.   They complain about ALSA yet they don't see adding more layers away from kernel is not improving issue.   ALSA also containes a sound server in libasound that also could be bypassed because its also in the wrong place.

One of the ways to screw Pulseaudio up massively is enable dmix in ALSA under pulseaudio so doubling up sound servers so leading to broken up audio.   Pulseaudio has made audio debuging more complex.   Should we be happy no.  We want simpler audio stack with less glitch causes.  Less glitch causes means don't stack the bugger like Pulseaudio has.

Pulseaudio is a poor design.  A echo of many poor designs.  How many times do we have to repeat.   You and thousands.  Ok thousands used artsd that is dieing and will disappear.  Thousands used esound that is basically dead.   Thousands uses NAS that is dead.    Sorry Thousands even millions means nothing long term.   If the design is bad it will die in time when something better comes along.   This is how open source works.  I have been around long enough to see sound servers come and go.

The question for us is not if Pulseaudio is dead man walking.   Its how long before we get to write up its tomb stone when something better appears.

I work with the audio team of the Linux Standard Base.  And sorry my statements hold weight for closed source applications and open source applications that are not distribution bundled and distribution bundled.  I most likely hold more weight in my own right than then any single distribution audio team.   What is selected by LSB sets advisable path for application developers.   Yes I effect upstream of Distributions.   Pulseaudio API are excluded from being able to be included into LSB due to its design issue.   Gstreamer is on the consideration list it don't have design issues.  

Don't be foolish again Oli Warner lot of us here are not low level people.

More information about the wine-users mailing list