Re: [SQU] Squid doesn't download updated sites

The problem you're referring to, to be annoyingly complete about it:

When Squid is acting transparently, Internet Explorer in every version
prior to 5.5SP1 (and for huge corporate customers using the NT version
of IE, there is apparently a hotfix for 5.01, also, which I have not
seen as I am not a huge corporate customer), will not send a no-cache
directive of any sort to the cache. It does this because it believes
that there is no cache between it and the origin server (a not entirely
unreasonable assumption, since transparent caching is technically not
entirely a compliant action).

With the ie_refresh option turned on every IMS request from a sub-5.5 IE
will be turned into a forced no-cache request. This will cause fresh
content to always be loaded on pre-5.5 IEs. On 5.5+ IEs (even ones
prior to the fix, since there is no way to tell which version is talking
to the cache) the cache acts just as it does with Netscape, and requires
a no-cache request to come in before it will force a refresh.

To force a refresh in IE 5.5SP1+ and Netscape (all versions) you must
force a refresh using either CTRL-F5 (CTRL-Reload) or SHIFT-Reload,
respectively. This causes the browser to skip it's own cached copy, and
any upstream caches that respect the Cache-Control:no-cache or
Pragma:no-cache directive in the request header, depending on whether it
is IE or Netscape doing the requesting.

Clear as mud? So there was no bug in the way Squid was handling things.
   But now there is a fix for the lack of a bug in Squid. ;-)

There is also a version of the patch for Squid 2.2STABLE5+hno on
my home page:

http://www.swelltech.com/pengies/joe/patches

Awais wrote:

> As far as I know...there is a documented problem with refresh and IE
> ewhich is resolved in 2.4-PRE-STABLE with the ie_refresh tag.
>
>
> Awais
>
>
> Robert Collins wrote:
>
>
>> Hi fabian, squid's refresh pattern is used when squid needs to
>> _guess_ the expiry time for an object. The pattern you quote
>> below of 0 20% 4320, for an object 1 day old will result in cache
>> hits for 4.8 hours. (20% of the age of the object).
>>
>> After 4.8 hours squid will send the request upstream as a IMS (if
>> modified since) request.
>>
>> The EXPIRES header simply lets sites tell squid rather than it
>> guessing.
>>
>> If hitting refresh in your browser doesn't work (this should get
>> squid to send a IMS request even if less than the hit time (in
>> this example 4.8 hours) has passed) then you may be getting bittn
>> by the following:
>>
>> There are issues with 'transparent' caching, if you are using that
>> then that is almost certainly the issue. Try configuring your browser
>> with the proxy set manually. You may find that fixes your issue.
>>
>> Rob
>>
>> ----- Original Message ----- From: "Fabian Krämer"
>> <fabian.kraemer@VITODATA.CH> To: <squid-users@ircache.net> Sent:
>> Tuesday, February 06, 2001 5:19 PM Subject: WG: [SQU] Squid
>> doesn't download updated sites
>>
>> Hi Martin,
>>
>> I don't blame squid, it's just a question. By the way: before I
>> posted this messages to the group a cleared the local cache of my IE.
>> The problem is still appearing! I guess I have this problem
>> because the provider of this site doesn't support the EXPIRE
>> variable for this site. So squid have to wait the MAX_AGE timer
>> (in this case 72 hours) until it's reloading the page.
>>
>> Greetings from Switzerland Fabian
>>
>> -----Ursprüngliche Nachricht----- Von: Martin A. Brooks
>> [mailto:martin@hinterlands.org] Gesendet: 02.02.2001 22:32 An:
>> Fabian Krämer; 'squid-users@ircache.net' Betreff: Re: [SQU] Squid
>> doesn't download updated sites
>>
>> At 14:40 05/02/01 +0100, Fabian Krämer wrote:
>>
>>> Hi,
>>>
>>> I've installed Squid 2.3 Stable 2. The problem I have is that
>>> squid doesn't download updated html-sites. We have posted
>>> modifications to a site hosted by a provider. If we want access
>>> this site by squid we can only see the old site (without the
modifications).
>>>

>>> It seems that squid looks for site changes only every 72 hours
>>> like set in "refresh_pattern 0 20% 4320". Squid
>>> doesn't notice that the sites were modified.
>>
>> Many browsers cache webpages independently of the Squid cache
>> itself. The worst culprit for this is IE, which often caches
>> pages despite setting it's cache size to 0, forcing (in theory) a
>> per-instance check and saying never ever /ever/ EVER cache
>> anything, /EVER/. :)
>>
>> It might be worth checking this out before blaming Squid. Check
>> it, /then/ blame Squid :)
>>
>> Regards
>>
>> Martin A. Brooks

                                    --
                       Joe Cooper <joe@swelltech.com>
                   Affordable Web Caching Proxy Appliances
                          http://www.swelltech.com

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html