In message <199512282117.QAA10129@jekyll.piermont.com>
	so spake "Perry E. Metzger" (perry):

> Our present syslogd specifies a timeout of 1 second on the lifetime of
> such processes, which is pretty low.
> 
> The 4.4lite2 syslogd has a lifetime in it of five minutes, which is
> pretty high.
> 
> These are rather different numbers (!) and I was wondering if anyone
> had an opinion on which is more appropriate.

I would say 5 min is way too high.  We've had problems with the
sunos syslogd becoming a fork bomb when a tty becomes "stuck"
(ie: writes block) and log messages come in quickly (several
per second).  The process table fills up rather quickly if the
problem goes unnoticed, and that is with a 30 second timeout :-(

I'd hate to see NetBSD afflicted with this problem as well.
Ideally one could use non-blocking I/O and avoid the fork()
completely.  No, I'm not volunteering, I'm busy enough as it is :-)

 - todd