i personally think KDE team really need to jazz up the sound base abilities and include a sound recorder to record particular virtual channels. This would shut up any short falls.
I'm particularly looking forward to alternate sample rates. Up till now, I have compromised by using 88.2KHz. It's close enough to 96KHz but also exactly double 44.1KHz so resampling CD audio doesn't require as much CPU power as 96KHz would. It did produce crackles in non-CD audio in earlier versions but that seems to be fixed now. Still, this new feature seems like a better approach so I'd like to give it a try.
The best news is that PulseAudio now has a fixed release cycle. Until now PA has a very unpredictable release cycle and a buggy release could have been the latest release for about a year.
If it means that PA will finally properly work on my system, I'm fully in favor of it.
I think there are no traces of PulseAudio in my Slackware distro so all programs are using Alsa directly. What critical features I am missing?
Finally, this was a real deal-breaker for me.fixed Hurd operating system support
ALi page but absent on the ASUS page. If I'd realised the ASUS card didn't have hardware mixing before I bought it, I might have gone for something else but PulseAudio sidesteps this issue as well as providing better sound quality and reliable handling of surround sound than ALSA with dmix does. Trust me, I was a PulseAudio refusenik in the beginning and I tried really hard to make dmix work but no amount of fiddling with asoundrc allowed me to mix surround sound with other sources. I also tried JACK, which was good except that it had no way of automatically upmixing stereo to surround. PulseAudio, on the other hand, worked great out of the box. It's also not the resource hog that everyone makes it out to be. Right now, I see that it is using a whopping 4MB. Wow. As for CPU usage, that is down to the fact that it resamples using a much better quality algorithm than dmix does by default. Ubuntu changed the PulseAudio default resampler for this very reason. dmix can also be configured to use libsamplerate instead, for a fairer comparison.