Re: ESI for squid

From: Joe Cooper <joe@dont-contact.us>
Date: Sat, 10 Aug 2002 00:56:02 -0500

Wow. You've been in stealth mode for this one Robert! (Ok, I saw the
CVS entry for the SPONSORS bit a few days ago. ;-)

Anyway, great to hear of it. Should allow a lot of cool new uses for Squid.

Robert Collins wrote:
> Zope Corporation - www.zope.com - are funding development of ESI -
> www.esi.org - for squid, with immediate inclusion into the GPL codebase.
> So the first thing is a public 'Thanks!' for their input into squid.
>
> ESI is an xml based language designed for in-intermediary page assembly.
> The idea is that dynamic pages are usually made up of many independent
> fragments that will remain (mostly) static, even though the whole page
> is changing all the time. (i.e. news article summaries don't change much
> once published, but the choice of which summaries are displayed will
> change as new articles get published). ESI looks to be of great use to
> web site publishers, with it's ability to move multiple page fragment
> retrieval out of the web server, and into the cache farm. See
> www.esi.org for more information.
>
> The first (and nastiest) phase is nearly complete, which is allowing
> squid to initiate internal requests that
> * Do not have a client connection
> * Are still Varied, redirected, and access controlled.
>
> Following this I will be moving onto the actual ESI parsing and
> processing, and finally the 'edge architecture' specification which is
> what triggers ESI processing, and controls some of it's optional
> behaviour.
>
> The current work will immediately make data manipulation/examination
> add-ons to squid much easier to code, via the clientStream concept. I'd
> like to get this work into HEAD asap, so if another core developer could
> give it a once over, that would be fantastic. This is a very small
> subset of the changes I implemented for the content processing branch,
> and avoids *all* of the range requests and server->store->client
> information loss issues that that branch ran into. They still need to be
> addressed at some point, but not for the ESI project.
>
> If anyone wants to chip in and play with, or contribute to the ESI
> implementation, or some sub-section of it (i.e. if you are developing
> something that needs to hook into clientSendMoreData, and you can use
> this core infrastructure..), that would be great - the more the merrier.
>
> The CVS TAG is 'esi' on cvs.devel.squid-cache.org. The project is also
> accessible via http://devel.squid-cache.org/projects.html#esi.
>
> Cheers,
> Rob

-- 
Joe Cooper <joe@swelltech.com>
Web caching appliances and support.
http://www.swelltech.com
Received on Fri Aug 09 2002 - 23:59:05 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:16:02 MST