Re: [squid-users] Squid-2, Squid-3, roadmap

Dodd, Tony wrote:
>> -----Original Message-----
>> From: Amos Jeffries [mailto:squid3@treenet.co.nz]
> <snip>
>> 3.0 was about parity with needs. It failed some in that regard.
>> 3.1 is about making up that failure plus some.
>> Is seamless IPv6, SSL control, and weighted round-robin not enough of
> a
>> killer app for you?
>>
>
> SSL control is nice, but we don't use SSL anywhere near squid, so it's
> not a big issue either way for us... we already have weighted
> round-robin in -2 by using CARP with specific weights, unless you're
> talking about something different?

True round-robin looping, but with peer weightings affecting the
load-balancing on every cycle. It's much simpler then full CARP meshing.
Being geared towards a two-tiered squid setup which need bottom-layer
weightings calculated in the top-layer instead of flat-mesh setups like
CARP.

> As for IPv6... eh, I suppose that'll
> be nice if IPv6 actually starts getting use sometime this decade.

Some of us do. At least 5% of the net and growing. When 3.x rolls out
the capability fully you can expect a small jump in www traffic over v6
without any user-visible changes.

>
> As Mark said, multicore would be quite awesome,

Truely. Care to sponsor any of the cleanups being done to -3 for this to
happen?
http://wiki.squid-cache.org/Features/FasterHttpParser
http://wiki.squid-cache.org/Features/LogDaemon
http://wiki.squid-cache.org/Features/NativeAsyncCalls
http://wiki.squid-cache.org/Features/NoCentralStoreIndex
http://wiki.squid-cache.org/Features/SourceLayout

> as would better memory management,

Tricky. Do we have an expert in the house who can do better without pay?
(nobody seems willing to sponsor any memory management)

> better I/O throughput on the cache_dir,

Work was sponsored to get COSS into -2 I believe to help with this. But
there were not matching changes sponsored for -3. Anyone?

> proper support for memory only caches,

http://wiki.squid-cache.org/Features/RemoveNullStore

> support for acl based cache_dir's (i.e. cache_dir
> foo allow dstdomain blah while denying everything else and cache_dir bar
> allow dstdomain boo while denying everything else) to improve overall
> hit-rate and decrease cache file flapping,

http://www.squid-cache.org/Versions/v3/3.0/cfgman/cache.html

> handling of a cache_dir
> failure that doesn't include squid dumping core,

?? details or bug IDs please.

> HTTP/1.1 support,

Underway in BOTH squid versions.
http://wiki.squid-cache.org/Features/HTTP11 (a little out of date now).

Each version has a different set of compliance/non-compliance depending
on which areas developers have worked on since we gave this emphasis.

> options support.

Please describe this?

>
> Things in 2.6 I'd like to see in 3 (on top of mark's list):
>
> COSS support - stable, with all the functions -2 has

Thanks. I've added it to the WishList.
Are you interested in sponsoring a developer to get it into the TODO
List with an integration time.
http://wiki.squid-cache.org/Features/COSS

> follow_x_forwarded_for

http://www.squid-cache.org/bugs/show_bug.cgi?id=1628

> refresh_stale_hit

Part of the collapsed-forwarding feature I believe.
http://wiki.squid-cache.org/Features/CollapsedForwarding

> umask support

Thanks you. I've added this as a bug item.
http://www.squid-cache.org/bugs/show_bug.cgi?id=2254

>
> -Tony

Amos

-- 
Please use Squid 2.6STABLE17+ or 3.0STABLE1+
There are serious security advisories out on all earlier releases.