Antwort: Re: Antwort: Re: Antwort: Re: Antwort: Re: Antwort: [Mod _gzip] Vary: header and mod_gzip

From: <Michael.Schroepl@dont-contact.us>
Date: Thu, 29 Aug 2002 16:57:19 +0200

Hi Henryk,

> What Squid can do is to have a blacklist where it mistrust the
> (possibly absent) Vary header sent by the HTTP server and forwards
> the request even if there is a matching reply entity in the cache,

this is what I was hoping to achieve, and what would have
helped the most popular browsers in Squid versions of 2.4
and lower.
It would have led to mod_gzip output not being cached by
Squid, i. e. some implicit blacklisting, under the assump-
tion that Squid would have used some "consistency equations"
like "if there is a content-encoding there _has_ to be a
Vary: header because we know this might easily overdo
UserAgent's capabilities."

> Another thing Squid might need to add is a blacklist of mod_gzip
> prior to the future version that makes use of Vary to prevent
> mod_gzip responses from being incorrectly cached.

This exactly was the reason why I didn't only patch the
"Vary:" header into mod_gzip 1.3.19.1b but changed the
version number as well.
If Squid will receive a response from mod_gzip 1.3.19.1b
(which would be part of the content of the "Server:" HTTP
header, unless of course the Apache configuration will
shorten this header to not display the modules ... - this
would then become another "don't do this as you would then
prevent the Squid proxies from working best") it would now
be able to know the reason why "Vary:" headers are created
and why they aren't, and the same for future versions.

Likewise, gzip_cnc (which is probably not fully compliant
to any HTTP level at all ...) is at least configurable in
a way that it send its own proprietary HTTP headers like
"X-Gzipcnc-Version" and a version number (and this is even
activated by default, but can be switched off by some flag
in the Perl source code configuration section - should I
perhaps rather use a "Via:" header, and with which content?).
So if anyone wants to blacklist my "compression cache"
(until its HTTP compliance is a little better), I'd rather
like this to happen than causing problems for other people.

Greetings, Michael
Received on Thu Aug 29 2002 - 09:05:55 MDT

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