> I'd guess that after incrementing the pool size the delayed connections are
> always handled in the same sequence so all the available bandwidth goes to
> the first connection and there never is any bandwidth left for the new
> connection(s).
Ahhh... well the file descriptor which comes first in the scan after the
select() will take all the bytes, yes. If that's always the same one, that's
the problem. It shouldn't always be the same one for general "fairness"
reasons.
David.
Received on Fri Nov 13 1998 - 01:44:50 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:43:02 MST