For the list:
Thread title is a complete lie. "Jobst" has an ongoing set of requests about turning Squid into a silent black-box relay for certain traffic so that the proxy admin cannot see what is going on in his own network.


On 09/02/11 14:49, Jobst Schmalenbach wrote:
Hi

How can I let packages/sites "bypass" Squid?

I do not mind if people listen to online stuff, what I mind is that I end up 
with loads of entries in the squid log and in the cache.

For example I want squid not to touch/log/cache/whatever any packet that is 
"application/x-fcs" (and other media stuff)


* squid will "touch" the HTTP portion of every object going through, no exceptions * logging omission you asked about two weeks ago and got an answer on how to hide the traffic from yourself
* shared caching omission is below


Is this correct, i.e. it will allow it through but not log nor cache it?

Also is my understanding correct that ACL are cumulative (as below) so I can 
use multiple lines for the same ACL name?


acl media urlpath_regex \.(afx|asf)(\?.*)?$
acl media urlpath_regex \.flv(\?.*)?$
acl media urlpath_regex \.swf(\?.*)?$

The above for a *set* of OR conditions. Tested individually in sequence. If any one of the three patterns matches the ACL name is match.


acl media rep_mime_type x-fcs

cache deny media

You have the general idea of how to prevent things being re-used form disk (a disk file will likely still be opened for backing the RAM in-transit copy).

There are two problems though:
1) each ACL name can only have one type. You need one for urlpath_regex and anther one for the rep_mime_type

2) the rep_mime_type being a *reply* mime type will not match on requests when decision is made whether to open a file and store the future data directly.

I'm still wondering though why you want to do this? all the media types which can be proxied by Squid are potentially cacheable for a great bandwidth/speed savings. The non-cacheable ones will get discarded anyway.

Amos
--
Please be using
  Current Stable Squid 2.7.STABLE9 or 3.1.11
  Beta testers wanted for 3.2.0.4

Reply via email to