ICP miseries

Hi all,

after hearing about misterious ICP problems on this list and experiencing similar ones,
maybe I found out something important:

bind uses separate UDP sockets to listen for each interface address:
udp 0 0 127.0.0.1:53 0.0.0.0:0 -- loopback
udp 0 0 193.91.87.5:53 0.0.0.0:0 -- ppp0
udp 0 0 10.1.2.3:53 0.0.0.0:0 -- eth0
and one for outgoing requests:
udp 0 0 0.0.0.0:53 0.0.0.0:0

and sends replies always on the same socket on which the query came in.

I guess, Mr. BIND (Paul Vixie) arranged so not without a reason,
but without knowing the reason, I propose to follow Paul.

If in doubt, ask him!

Just to be sure:

Did everybody, experiencing those problems, have multihomed hosts involved?

Andrew. (Endre Balint Nagy) <bne@CareNet.hu>
P.S. As I remember, rwhod does it in a similar way. Maybe the trick is older?
Received on Sat Aug 31 1996 - 14:46:10 MDT