Re: [squid-users] squid 2.7 vs 3.x (was: Re: [squid-users] squid-2.7 pre-release testing)

From: Adrian Chadd <adrian@dont-contact.us>
Date: Mon, 14 Jan 2008 11:48:58 +0900

On Mon, Jan 14, 2008, Robert Collins wrote:

> > 3.x: has some internal code restructuring, is a C/C++ hybrid, includes
> > integrated ICAP support; Amos has ipv6 support included in 3.HEAD.
>
> IIRC tagged delay pools were merged, definitely per-user delay pools
> (class 4 pools) support was merged, ESI, modular disk IO stuff taken
> further.

ESI is still not production ready, the modular disk IO stuff is a good
start but there's still legacy code (src/disk.cc) which implemented
the whole original "modular" code (sync or pthread-aio) which was
made less useful when the store io stuff was split back in 2.3/2.4 days.

> Some of that is really nasty to try to get working in 2.x (in particular
> the tagged delay pools), and is largely benefits from 'internal
> restructuring'.

Yes, but the internals are still ugly. I don't think the current state of
the delay pool "hooks" into Squid, especially in the comm mess, is actually
the way to go. Unfortunately there's little motivation to fix up this
stuff.

Adrian
Received on Sun Jan 13 2008 - 19:39:27 MST

This archive was generated by hypermail pre-2.1.9 : Fri Feb 01 2008 - 12:00:04 MST