W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > September to December 2000

RE: caching problems

From: Joris Dobbelsteen <joris.dobbelsteen@mail.com>
Date: Thu, 26 Oct 2000 13:59:26 +0200
To: "'Tim Coates'" <tcoates@dynamics.net>
Cc: "WWW WG (E-mail)" <http-wg@cuckoo.hpl.hp.com>
Message-ID: <000301c03f44$309dfc40$01ff1fac@Thuis.local>
> -----Original Message-----
> From: Tim Coates [mailto:tcoates@dynamics.net]
> Sent: Thursday, 26 October 2000 3:51
> To: http-wg@cuckoo.hpl.hp.com
> Subject: caching problems
> 
> 
> Dont know if I am doign this the right way?!?
> 
> I am wondering what happens when HTTP 1.1 rules are specified 
> in an HTTP 1.0
> protocol - particulariy in regard to Cache-Control and proxies.
> 
> Here is the environment.... IIS server, MSIE 5.0 browser and 
> Squid Proxy.
> 
> I have a web page that should not be cached by a proxy/browser etc.
> Appropriate Cache-Control headers are sent also. In the GET/POST and
> responses, the HTTP version is specified as 1.0. (Yet HTTP 
> 1.1 Cache-Control
> rules are also used.)
> 
> I have two versions of the same web software. At the browser 
> I am able to
> retrieve a web page from cache (one time only!) whereas with 
> the latest
> version, the page does not appear to be cached as the web 
> page is retrieved
> from the server. (This is proven using a packet sniffer.) In 
> both versions
> of software, packets sent by the server contain no-store, no-cache,
> must-revalidate, max-age=1 cache control directives. The only 
> difference
> between the two version of the software is that in the latest 
> version, the
> headers are sent separately to the web page. In the earlier 
> version, part of
> the document is sent with the headers.

You are using too many cache-control directives. no-cache is sufficent..
No-store is also enough, causing it not to be stores anywhere.

no-store I would only recommend for sensative data, no-cache for non-sensative data, because it can save some network bandwidth.

using also the must-revalidate and max-age may confuse the cache, as it is *probably* non-compliant.

> 
> I am aware of the reliability of caches with 1.0 protocols, 
> but if the only
> difference (after using a packet sniffer) is that the headers are sent
> separately, then I am at a loss to determine why in one case 
> a page is be
> retrieved from cache (previous version) whereas (in the 
> latest version) the
> user is redirected to a separate page.
> 
> 

Maybe that you should also include

Pragma: no-cache

just in case. Using HTTP/1.1 is better....

I expect a HTTP/1.1 cache to interpet the cache-control header included in a HTTP/1.0 response. MSIE does this, I think, but including "pragma: no-cache" is much more reliable...


- Joris
Received on Thursday, 26 October 2000 13:01:34 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:33:40 EDT