9 Apr
2005
9 Apr
'05
2:10 a.m.
On Fri, 2005-04-08 at 23:23 -0400, Jean-Marc Valin wrote:
I respectfully disagree with this interpretation. Everyone on LKML
seemed to agree that Con's scheduler was the best solution proposed, but
will have to wait for 2.7 as it's too intrusive for 2.6. The only
question now for 2.6 is whether the realtime LSM or the rlimit based
solution will be accepted.
Personally I prefer the realtime LSM but would accept the rlimit based
approach. It puts more of a burden on the distro/sysadmin to configure,
as it requires patching PAM and bash and a correct limits.conf
configuration. But from the users point of view the end result is the
same (configure a "realtime" or "audio" group and add users to it).
What I don't understand is that the people that the kernel developers
*do* listen to by default aka the big distros like Red Hat and SuSE, did
not weigh in sooner on this issue. I suspect that specialized needs
like low latency audio are not too high on their priority list. The
"desktop people" at Red Hat still seem to be focused on making the basic
user experience compare favorably to Windows and Mac OS - if the average
laptop can't even boot FC3 and have all the hardware work, they can't be
too worried about the 0.1% of users who want to use their laptop for
live performance.
Lee
If being an
audio developer carried any weight with these people,
you'd think they would listen to Paul Davis or Lee Revell or me. They
don't.
Then, they might listen to Paul Davis, Lee Revell, you, me, authors of
Vorbis, Theora, GnomeMeeting, and so on at the same time.
Con Kolivas is one of the good guys. But, the
"arrogant fools" don't
listen to him much, either.
As far as I see, Con is mostly having problems showing that there's an
interest. The fact that I'm the only person that tried his new scheduler
didn't help.