At 4:19 PM -0700 4/18/01, Roy wrote:
> I changed the -T setting to 300 seconds and was still getting
> timeouts. Since I
> regressed back to version 3, the timeouts ent way for the most
> part. I think
> something is broken in the timeouts. It might be if a single
> message takes more
> than the -T value, it times out.
If you're getting 'POP Timeout's, that means Qpopper failed to
receive input from the client within timeout seconds. The timeout
counter only starts when Qpopper is waiting to receive input -=-
it's unrelated to sending data.
If you're getting client timeouts, that means the client
disconnected because it didn't get within its timeout period.
In either case, if these occur more frequently with Qpopper 4.0, it
may be a combination of congestion in your network, small timeout
values in the clients, and Qpopper 4.0's default behavior of
aggregating small network writes into a large one. You can try
adjusting 'chunky-writes' and see if this makes the problem go
away. See the FAQ entry at
<http://www.eudora.com/qpopper/faq.html#timeout>. Note that you
can set chunky-writes per-user if you like. This is also discussed
in the Administrator's Guide (the GUIDE.pdf file in the
distribution) in the "Performance" section.
