Re: Querystring vs. Squid Cacheserver

I still dont see the point. The issue is whether or not the script gives
the same output given the same input at a later point in time. Even though
you are right that words like 'sex' tends to be often used when querying
search-engines, there are no way a cache can know that the output wont be
changed due to adding or deleting of urls in the database, conditioning
formatting of output depending on the browser used and so on.

Since this is the case in 99% of the urls including a question mark I will
keep my squid.conf as it is - rather than pollution my cache with 99% junk
just to respect (your) 1% cache-friendly urls.

I would suggest that you instead of your current approach made a handler in
apache's srm.conf (eg ScriptAlias /citymap /cityscript.pl). In the I would
then parse the REQUEST_URI for the parameters. That way you can make the
script look as a static page by referring to it as
http://www.foo.de/citymap/berlin/whatever.html. Then your headers would be
respected, but you should then also add features to your script that makes
it possible to validate if the 'page' has been changed (ie handling of
IMS-requests and add an E-Tag).

Before you do that I still think that you should consider what the
likelihood is that the same 'page' will be shown twice within your
expire-time. If it is unlikely I dont think it is cache-friendly to make
every cache in the world cache a copy of your 'page' just because it is
possible.

Written by you 11:41 02-03-99 +0100,Andreas J. Koenig
>>> If people installing squid really follow that recommendation, all my
>>> careful considerations about Last-Modified and Expires headers are
>>> worthless because *all* my URIs contain questionmarks.

>I'm talking about a thing where a cache server makes a lot of sense
>and _I_ do behave cache-friendly. But if cache maintainers ignore the
>headers just because there is a question mark in the URL, they behave
>not cache-friendly. That's not rational, so I believe, the
>recommendation in the squid.conf is bogus.

Regards

-- 
Ole Møller olm@cybercity.dk, Sysadm CyberCity Internet
Micro$oft is not the answer, Micro$oft is the question, the answer is no.