On Thu, Jun 20, 2002 at 12:06:20AM +0200, Manuel Bouyer wrote:

> No, trust me this happens with 1.5.2 too. I've got to hack pop3 and imap
> daemons to make it acceptable.

This is very true.  I suspect UBC has made things a tad worse in certain
cases.

The only real way to gain more insight is for a lot of people to run
1.6 on various workstations and servers, and report their results.
IMHO, release it, and use the feedback to set the course for 1.6.1.

That's the way it's supposed to be done, barring truly bad behavior
with 1.6.  What I see so far is people on both sides, not a definite
leaning one way or another.  So it seems reasonable to use 1.6 release
as a shakedown.

> No, the root problem is that a single, small I/O can need several seconds to
> complete because some other fills in the I/O queue.

Sounds like your garden variety resource starvation problem.  

Isn't the I/O queue processed in a way that limits heavy I/O processes
to give better response time for infrequent I/O processes?  If it is,
is this tunable?


-- 
UNIX/Perl/C/Pizza__________________________________shannon@widomaker.com