Re: [squid-users] Vary / Accept-Encoding / mod_gzip Redux

From: R Pickett <emerson@dont-contact.us>
Date: Wed, 7 May 2003 16:54:26 -0700 (PDT)

On Thu, 8 May 2003, Henrik Nordstrom wrote:

> Quick check: What is the HTTP headers of the reply you think is wrong?

Apart from the GET example? The GET example is basically the behavior I'm
worried about, getting back cached gzip content when there's no Accept-Encoding
header.

> PURGE only deletes one single object variant, not all object variants
> of this URL.
>
> To PURGE a specific object variant you must include the request
> headers matched by Vary. if you wish to delete multiple variants
> including the variant with no matching request headers then purge the
> one with no headers last.
>
> If no request headers is included then you will purge Squid's memory
> of what the Vary header looks like for this URL. As soon as there is
> one request for this URL Squid will learn the Vary header again.

Oh, that's a very bad misunderstanding on my part then, and is a bit of a
problem for our application. Is there any sane way to tell squid to purge a
URL's full list of variations without knowing what they are ahead of time?

Alternatively, is there a deterministic way to get a list of variants from
squid for a URL? We cache rather aggresssively, and we want to be able to tell
squid to drop its current copy, programmatically, when we update our backend
database for a URL.

> This is odd. It sure looks like the first Vary response above. Should
> not happen, and is not what seem to happen in my tests. Here
> everything seems to work fine.
>
> Can your try to repeat this with "squid -k debug" enabled and then
> send me the results in cache.log?

I'll do that under separate cover, direct to you so as not to spam the
entire universe with that.... Thanks tons for looking at this.

--
R Pickett <emerson@craigslist.org>
craigslist.org
Received on Wed May 07 2003 - 17:55:02 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:16:21 MST