httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Andrew <and...@aaaaaaaa.demon.co.uk>
Subject Re: proxying
Date Mon, 22 Jan 1996 20:41:39 GMT
Paul:
> I discussed this with Andy some time back, http seems to be lacking a 
> mechanism for putting a ttl on pages so that caching mechanisms can be
> written that allow servers to provide pages from a cache and then be
> forced to retrieve an authoritative copy when the ttl times out. From his
> description the current cache related headers don't seem to work in
> that way. It would be something like the way DNS handles it.

Woah, hang on there tiger-boy.  There is an Expires: header which
relates to the time after which a caching-proxy (like harvest, CERN
or even Alexei's module) would be *advised* to reload any stale files.
It seems only to relate to the relationship between source servers
and caching-proxy servers however, browsers aren't intendied as
recipients of Expires: headers, on my meagre reading of the subject.

It's used as a way of allowing the source site to influence the
eventual expiry date of old data (akin to TTL) rather than relying
on the good management of proxy servers, regularly trashing old data.

Apache doesn't issue Expires: headers and since we've not approached
Apache's construction in a cache aware way it's hardly suprising.
CERN *is* able to issue Expires: per-file (and presumably by-type)
by action of it's *.meta file syntax, implementation of which would
make a cool module incedentally.

Whatever, the 'intended' semantics of Expires seem to be less well
understood than more widely used aspects of HTTP, it does bare more
looking into I think.

>   Paul Richards. Originative Solutions Ltd.

Paul, when ya gonna show me how to rebuild this damned kernel, anyhow?

Ay.

Mime
View raw message