Re: [squid-users] squid / DNS name servers

On Sat, Sep 09, 2006, Odhiambo WASHINGTON wrote:

> | If Squid doesn't do it then please file a feature request
> | in Bugzilla and we'll get to it after we've released Squid-3.
> | Yes, we're going to release Squid-3.0. And then 3.1. Honest :)
>
> You are going to release them even before the dust settles for 2.6? ;)

The dust has pretty much settled for 2.6. No new features are going
into 2.6; its mainly bugfixes. Quite a bit of work has gone into
the cyclic filesystem (COSS) by Steve Wilton but the rest has been
fixing whatever pops up in Bugzilla.

> BTW, why 3.0 and 3.1, if you're honest enough?

We (squid core) are working on a roadmap to outline what our goals for
the ongoing project are and what we're looking to achieve for each
release.

So far, we've agreed that 3.0 will be released after we've fixed the
current set of bugs. 3.1 will be released after some more code
restructuring and performance work has been completed.
These goals haven't been "fixed" yet and may change before we announce
the official roadmap.

Now, why 3.0? A few reasons:

* Although some of the changes in 3.0 are a bit "exciting"; the change
  to C++ and the ongoing work in 3.0 and 3.1 will result in a much
  cleaner codebase; much easier to implmement new and strange features
* The client streams model in 3.x will (and does!) allow for content
  manipulation between request/reply and to/from the storage manager.
  There's an example module that implements gzip compression. Other
  modules could include HTML content rewriting (such as what ESI does),
  image re-encoding, in-line virus/spam scanning, etc.
* The 3.x branch is closer to HTTP/1.1 compliance. Its still quite a way
  off; but there's hope for it in 3. :)
* 3.x has functional ICAP support. I've been told it works rather well
  but I haven't used it myself.
* Robert Collins has been working on a -lot- of unit testing. The idea
  there is to define an API and behaviour and test code to make sure they
  implement the desired behaviour. Its quite difficult to test parts of
  code in the 2.x branch - everything seems to depend on everything
  else :)

Adrian
Received on Sat Sep 09 2006 - 04:54:25 MDT