[squid-users-request@ircache.net: Re: squid problem]

I've joined the nlanr cache hierarchy with my squid proxy to save on
bandwidth and reduce latency, but squid seems to be doing some stupid things
despite ICMP support being turned on:

24.128.178.0 is my local subnet.

24.128.178.0 3/ 3 0.0 1.0 osiris.978.org ians.978.org
 www.978.org www.978.ml.org oniongirl.978.org vertex.978.org dennisr.978.org
 members.978.org 978.org www.osiris.978.org www.osiris.ml.org torgo.978.org
 osiris.ml.org
rtp.cache.nlanr.net 44.5 12.0

As you can see, the direct route to these sites is one hop, with a zero ping
time. The route via rtp.cache.nlanr.net is 44.5 and 12 hops. My question is
why does squid contact these hosts via the rtp.cache.nlanr.net proxy instead
of directly? This also happens for hosts on other closeby subnets that
should be contacted directly instead of via a parent.

I'm using Squid Cache: Version 2.1.PATCH2 on a Debian Linux system with the
2.2.2 kernel. I can provide my full config file if it's neccessary, but
here's a quick summary of the options I think are pertinant:

cache_peer rtp.cache.nlanr.net parent 3128 3130
cache_peer bo.cache.nlanr.net parent 3128 3130
query_icmp on
minimum_direct_hops 4

-- 
Brian Ristuccia
brianr@osiris.978.org
bristucc@baynetworks.com
bristucc@cs.uml.edu
----- End forwarded message -----
-- 
Brian Ristuccia
brianr@osiris.978.org
bristucc@baynetworks.com
bristucc@cs.uml.edu