Re: Squid 3 features

On Wed, 2003-05-07 at 10:00, Henrik Nordstrom wrote:
> On Wednesday 07 May 2003 00.45, Robert Collins wrote:
> > Ok, are there any features that folk want in squid-3 that aren't
> > there already?
>
> "overlapping" helper lookups. I think we should have this at least for
> external_acl and other simple helpers, but preferably also for
> stateful helpers.
>
> The patch today only exists for 2.5, and is not fully working for
> stateful helpers.

Do you have a timeframe for completion of this?

> I'd also like to se a cleanup of certain squid.conf directives which
> is a bit too arcane today.
>
> One example is reply_body_max_size where size factor should be added
> and the allow/deny column deleted, but there is more examples of
> these odd things.

Hmm, the reply_body_max_size allow/deny column is very useful: it allows
short circuiting the acl's for unlimited use - I'm not sure I agree on
this one.

> Another is always/never_direct. A single acl driven tri-state
> directive (direct,auto,parent) doing the function of both is more
> natural I think.

I agree. Again - expected timeframe?

> A third is request routing among multiple cache peer.
> cache_peer_access makes sense to programmers who have got stuck in
> logics and boolean thinking, but nobody else. Also I imagine the
> current scheme gets very inefficient if the number of peers and
> access rules are large.
>
> A review of the squid.conf syntax is needed to identify more of these
> stupid things.
>
> The change of major version is a very good opportunity for making
> cleanups like this.

IMO we'll be heading for unneded problems if we clump such a review in
with the already major changes. We can easily do backward compatible
changes as we move forward - I've been working on making that easy to
code in some development branches.

Rob

-- 
GPG key available at: <http://users.bigpond.net.au/robertc/keys.txt>.