Re: [squid-users] squid and long URLs

On Mon, 13 Dec 2004, Daniel Graupner wrote:

> No, my squid is inside a local network so it can only reach webservers inside
> this network. To access the internet (e.g. ibm.com) it has to use a peer
> which is also inside the network.

Then you are per definition inside a firewall and the FAQ applies to you.

> OK, I have to investigate why squid tries to go direct in my case.

Simply because you have not told Squid you are inside a firewall.

Squid is desinged to participate in a cache mesh (number of cache
servers talking to each other in various manners, parents, siblings,
childs, etc) where the purpose of the cache mesh is to gain a good hit
ratio beyond what a single server alone is capable of.

Because of this simply adding a cache_peer does not force Squid to use it,
it just tells Squid it may use it when Squid seems it is fit to be used
for the request.

In HTTP there is a large number of different types of requests where it
can immediately be determined the result almost certainly won't be
cacheable or otherwise won't be any caching benefits of forwarding the
request to a peer. On such requests Squid prefers to go direct when
possible, simply because it is more efficient than giving the peers more
work for things which won't benefit the hit ratio.

> yes i have the (standard) no_cache directive in my squid.conf

Good, and they should be there. This is not the cause to your problems.
Being inside a firewall without telling this fact to Squid is the cause to
your problems.

> K, I found the strip_query_terms directive, now its clear. For this one has
> to read the default conf, it is not mentioned in the userguide nor faq ;-(

Users are supposed to read the documentation included in
squid.conf.default. It is after all the authoriative source on how to
configure Squid. The FAQ and userguide are supplements to clarify what is
not easily understood from squid.conf.default.

Regards
Henrik
Received on Mon Dec 13 2004 - 09:18:22 MST