[Mod_gzip] "mod_gzip_send_vary=Yes" disables caching on IE? (1.3.26.1a)
Slava Bizyayev
mod_gzip@lists.over.net
Fri, 6 Dec 2002 08:55:44 -0600
Hi Jordan,
There is nothing bad with Vary header itself, the problem might be with
features of implementation of this header. I remember one hot discussion on
mod_perl mailing list about a year ago. That was very similar issue with
mod_deflate. I was lucky to demonstrate that Vary works fine (with M$IE)
even when you need to cache content locally. Apache::Dynagzip is using this
approach effectively.
Thanks,
Slava
----- Original Message -----
From: "Jordan Russell" <jr-list-mod_gzip@quo.to>
To: <mod_gzip@lists.over.net>
Sent: Friday, December 06, 2002 2:14 AM
Subject: [Mod_gzip] "mod_gzip_send_vary=Yes" disables caching on IE?
(1.3.26.1a)
> Hello,
>
> (Please forgive me if this has been discussed before; I did check the
> archives but nothing turned up.)
>
> While looking through my http access logs today I was alarmed to find a
lot
> of repeat requests for the same .gif files from the same hosts in a very
> short period of time. As it turns out, Internet Explorer (6.0, at least)
> doesn't want to cache objects that have a Vary header, so the same images
> were being reloaded on each and every page visit. I did some searching and
> found this:
>
> http://groups.google.com/groups?selm=uTYe1IjgCHA.2324%40tkmsftngp08
> "IE completely refuses to honor this VARY header and fetches new page
> each time regardless of Cache-Control directive if it encounters VARY
> header in server response. Now, RFC compliant browser (like Mozilla)
> sees that header and make separate requests for separate VARY headers
> specified)."
>
> Changing mod_gzip_send_vary to No solved the problem, but of course that
> creates issues with Squid...
>
> This seems like a significant 'gotcha' to me, as it can defeat the purpose
> of using gzip compression in the first place. If you have a site with
static
> pages and a lot of repeat visitors using IE, the total bandwidth consumed
by
> "mod_gzip + no caching" could end up being higher than "no mod_gzip +
> caching".
>
> Am I the first to notice this? Can anything be done about it? (Maybe
> mod_gzip should avoid sending Vary headers unless it detects signs of a
> proxy server?)
>
> Thanks,
> Jordan Russell
>
> _______________________________________________
> mod_gzip mailing list
> mod_gzip@lists.over.net
> http://lists.over.net/mailman/listinfo/mod_gzip
>