phoenix said:
Considering OSS on FreeBSD gives you almost all the features of ALSA+PulseAudio on Linux, why would we want to downgrade to ALSA?
My first FreeBSD installatiom is from today. I didn't get the time to look at the soud. Compared to ALSA+JACK, how do OSS on FeeBSD?
The main features of JACK are a constant sound latency (pulse is not abble of that and never will) and synchron operations of all the clients.
phoenix said:
Considering how many Linux developers are afraid of using/fixing ALSA directly, and instead fix issues with ALSA higher up the stack (hence the existence of PulseAudio), why would we want to downgrade to ALSA?
That's not true. Many of the developer's on the LAD are the same than the one on jack-devel, and they care about ALSA+JACK, which is a professional solution. The problem with the sound you are talking about is explained here:
RealtimeKit and the audio problem. All the story with pluseaudio begun with Lennart Poettering. It is 2 existing possibilies to get realtime sheduling on linux, one is the rt patch, the seond one is with RT_GROUP_SCHED and the cpu cgroup. The advantage of that second solution is you can get realtime scheduling with a vanilla kernel. He is also the guy behind *kit and systemd.
What is very funny is with a few line into my
~/.asoundrc file, it is possible to interface
ALSA and JACK, that with no mesurable added latency, and each ALSA application is visible in the graph of
qjackctl. So, ALSA and JACK is a much better solution than pulse and ALSA.
Systemd, when installed, will comppletely take over the control of the cgroups, and instead of implementing mean to make policies, which will give a great freedom to the users, it is implementing the policy, which cause already a bigger problem than the one systemd is intended to solve, and only about half of the job they want to do with it is done... No coment :beer For now, my USE flag on gentoo are "-systemd -consolekit -policykit -udisks -udisks2 -pulseaudio", but I don't know how long that will be possible. Maybe I am too pessimistic. It is not urgent, but I am looking for an alternative, so I try FreeBSD.
Another big issue on Linux is wayland that come. That will break a lot of applications, because the compatibility layer will never be complete. X and all its extentions are just too complex. The goal of Wayland is to get more integration. I think they missed an important point about integration. The best possible integration I know, is like with AROS, to integrate every thing into the graphic server, from the window manager to the toolkit. The result is that the Gimp in its hosted AROS version start at least 5 times faster than the native gentoo version. They can say what they want about the multicore processors and a modern GUI, but 1 core is more than enough to run a well writen GUI that doesn't interfere with the system.
And all these issues are related, because this is the same peoples pushing to rewrite the kernel cgroups, stuffs like pulseaudio, *kit and systemd, and wayland. For me, that will be a perfect OS for old generation mobile phone, because a trend is slowly emerging on the mobile market, this is multi-core processors with dedicated cores. 1 low power and low frequency core for the stand-by, one dedicated for the hardware, one general purpose for the GUI, and 1 DSP for all the calculations. I really hope that trend will reach the desktop market as well. That because DSP algorythms are well known and have proved to be much more efficient than any well optimised code on a general purpose cpu.
EDIT: typos
