Re: Active proxy 'problems'

Pre-fetching proxies are a Bad Idea, IMO (and many others'); they have a
number of unpredictable side effects (like this one). Nevertheless, some
products do it; Cacheflow [1] is the worst offender (because they do it
out of the box); IIRC Microsoft ISA server [2] can be configured to do
this as well.

Also, a number of browser "add-ons" do this for individual client caches
(mostly Windows).

There really isn't anything you can do to stop them, AFAIK; you might want
to write to Cacheflow and find out if they have any Cache-Control
extensions to disable this behaviour (please tell us if you have
success!). Another approach would be to have your Squid make its responses
uncacheable (by rewriting the outgoing headers); however, this would
disallow all downstream caches from keeping your content, which may use
more bandwidth in the end.

Cheers,

1. http://www.cacheflow.com/
2.http://www.microsoft.com/isaserver/





----- Original Message -----
From: "N. Coesel" <nctnico@cistron.nl>
To: <www-talk@w3.org>
Sent: Monday, July 01, 2002 10:12 AM
Subject: RE: Active proxy 'problems'


>
> At 09:06 01-07-2002 -0700, you wrote:
> >
> >>
> >> This setup seems to work quite well, but there is one major
> >> drawback: There seem to be active proxies at the client side.
> >> This means that more and more sites will refresh their caches
> >> every 24 hours. This results in a huge amount of wasted
> >> bandwith since no-one (= a person) at the client side
> >> actually requested the document.
> >
> >What do you mean by 'active proxy'? Is that something that pre-fetches
> >documents just in case a user may ask for it?
>
> Yes.. sort of. I've noticed that some proxy caches re-fetch (yes
re-fetch)
> the documents every 24 hours (or with shorter intervals) to keep their
data
> up-to-date. This results in a huge amount of hits.
>
> >Do you know what user-agent is provided in those requests?
>
> I'm recompiling squid (the fron-side proxy) right now to log the user
agent.
>
> Nico Coesel
>

Received on Sunday, 7 July 2002 21:17:53 UTC