>On Tue, 2003-09-02 at 09:20, Joshua Brindle wrote: >> Ok, after looking at LOTS of logs and tcpdumps of the traffic >> i finally figured out that when pages or images are broken there >> is an extra crlf after the headers. this only happens when using >> a cache peer, and happened with both dansguardian and tinyproxy >> as the cache peer. I'm using these cache peer options: >> allow-miss default no-digest no-netdb-exchange no-query name=localhost >> >> Since this does not happen when squid is run standalone it seems >> that the offending code is in the peer area, but I can't find anything >> that would add a crlf anywhere except client_side.. any help/pointers/ >> realizations? Thank you very much.. >> > >Ok, this is something I can look for - will get back to you in a couple >of days. > >Thanks for the excellent detective work. >
To update this problem, it does _NOT_ occur when i put another squid on the other side of the squid 3.0. For example, the setup that would not work was squid 3.0 -> dansguardian -> tinyproxy (the tinyproxy was to eliminate the routing loop as a possible source of the problem). However, if i put a squid 2.5 with icp between squid 3.0 and dansguardian ala squid 3.0 -> squid 2.5 -> dansguardian -> tinyproxy it works fine, although admitidly slow. So the problem must occur when using a cache_peer which does not have icp of digest query capability, although I simply can't find it. Joshua Brindle UNIX Administrator Southern Nazarene University
