Re: Querystring vs. Squid Cacheserver

>>>>> On Tue, 2 Mar 1999 16:06:29 +0100, Steffen Ullrich <ccrlphr@xensei.com> said:

> my 4 cents...
> 1) It definitly makes sense to specify Last-Modified and Expires from a
> CGI-Script. Take something like freshmeat.net (I don't checked if they
> use the headers). It would make sense to set Expires to smthg like 1 hour
> and Last-Modified to the time when the database was last time modified. Then
> the content could be cached from proxies which takes load from the
> server and also shortens response times for the user. Same true for a
> web view of a discussion list where a few times a day new articles come
> in and you offer various views (subject/thread/subthread). It doesn't
> make sense to recreate all possible views to a part of the list every
> time a new article drop in. Instead create them on demand (and probably
> cache the output locally too for a while) but tell the client about
> Last-Modified.

Hey this answer alone is already more than 4 cents worth :-)

> 2) To work around the problem with not caching '?' (and with the problem
> that some users find URLs like this strange looking) I usually use
> PATH_INFO and append the parameters to the script as a path (smthg like:
> http://server/script/view-as-thread/msg004). If the request is a result
> of the user filling out a form I sometimes redirect the user based on
> the input to a better locking (cachable, but dynamically created) page.

But PATH_INFO ruins relative URLs.

> 3) Why do you think that http://server/script?A=10&B=4 has to be the
> same as http://server/script?B=4&A=10 ?? Probably your script does it
> this way but I know of any rule who enforces that behavior.

Correct, that was a big thinko on my end. Thanks for spotting.

> 4) I think URL's with '?' in it should be cachable (but maybe set
> Expires to 0 unless a Expires header is given). For contents which
> should not be cached POST requests or a proberly defined Expires header
> should be used

Agreed.

-- 
andreas