[squid-users] Cacheboy Squid information

Hi everyone,

Just a fore-warning: this announcement has nothing to do with the
Squid project.

I've been in the position of maintaining some Squid-2.5 caches for a while.
My biggest problem: having to maintain a local branch (it was, for a while,
a few separate branches!) of squid-2.5 with relevant features (such as Linux
epoll()) to keep things running smoothly and effectively.

I've seen a few people try to do something similar and get frustrated.
I decided to setup a Monotone repository to work with one of said people
(Steven Wilton) to see how easy it'd be to collaborate on work and run
a single unified tree with the changes relevant to us. We've found this seems
to work well; we've been able to fix a lot more bugs between us than we
ever would have working with separate source trees.

We've decided to release all of this to the public. The hope is that others
who run Squid-2.5 installations with the patches available at
http://devel.squid-cache.org/ (such as the logging patch, the rproxy support,
etc) can use this source tree and provide bug reports and share feedback.
This way we all benefit from a single codebase, rather than everyone trying
to maintain their own patch sets.

The source downloads and project information can be found at
http://www.cacheboy.net/ .

The current state of affairs:

* WCCPv2 support; including support for defining multiple routers and multiple
  static/dynamic service groups
  (WCCPv2 MD5 authentication is in testing at the moment)
* Connection-pinning support: allowing sites using NTLM and similar authentication
  mechanisms to work correctly through a proxy
* Linux TPROXY support: spoofing the client IP address so the client and server
  don't "see" the proxy server sitting in between (which works with WCCPv2)
* Linux epoll() support

We're confident that our current codebase is as stable as Squid-2.5 is,
including the current features. Things are still at a pre-release stage
mainly due to documentation (and the lack of it) - but we hope that feedback
will help us solve that problem. The codebase is currently running on a number
of busy forward caches serving thousands of clients.

Again, none of this is endorsed by the Squid development team. This is something
I've decided to do on my own.

Adrian
Received on Mon Apr 24 2006 - 01:58:16 MDT