[KLUG Members] Reports from the bleeding edge....

Adam Tauno Williams members@kalamazoolinux.org
Wed, 2 Apr 2003 10:54:02 -0500


>>I think in RH8, yes.  All I remember is that installing 8.0 on my wife
>>laptop, sound just worked; as it does on iPaqs.  Neither was true with
>>7.x, that required dinking.
>I don't know if OSS is that much worse than ALSA at autoconfig. ALSA has
>built-in OSS emulation, try `ls -l /proc/asound`.

Yep, needed OSS emulation for Loki games to work on ALSA system.  It wasn't
really autoconfig that was better, just hardware support for newer chipsets such
as those founf commonly in laptops (iPaqs are really essentially stripped down
laptops in a desktop case).

>>>New POSIX Threads for Linux/Next Gen. POSIX Threads - I think both
>>>require large glibc changes and are still 2.5 only, though they are
>>>not actually included.
>>This is apparently in 9.0,  although I don't recall seeing it in
>>Pheobe (I run lots of Java apps, not of them crashed, a solid
>>indicator that POSIX threading wasn't there).
>Hmmm... I would seriously doubt it's in 9.0:
>http://www.redhat.com/about/presscenter/press/2003/press_rhl9/
> - First introduction of new threading technology: NPTL (Native POSIX
>   Thread Library)
>Holy crap, I could be wrong! Wow. (Course, I got half the name wrong.
>:-) <droooolll....>

Pheww, someone else is wrong for once, nice feeling.

>Hmmm... with the 2.4.20 kernel. They had to backport a metric shitload
>of 2.5 patches to achieve the high-efficiency kernel threading to

And rumor is that is why they yanked UML.

>support NPTL.  Bets say that 9.1 (or 10) will have 2.6.x.
>NPTL... must resist... urge to... convert....

Nah, convert now.  With this (like AIX) I will be able to instance more than one
virtual processor per physical CPU attached to the Informix shared memory
segment, a nice performance kick on DSS type queries (I assume the same will
apply to Oracle and DB2).  *AND* one won't need to throttle OpenLDAP slapd
threads - currently anything over like 12 threads causes the system to start to
stutter, so one artificially limits them with the "threads" parameter (And nscd
just started to resemble something stable, sheesh).