Re: minimum-cache time?+

Stephane Bortzmeyer wrote:
>
> On Sunday 27 October 96, at 16 h 6, the keyboard of "Miguel A.L. Paraz"
> <map@iphil.net> wrote:
>
> > acl noncacheable .microsoft.com
> > must_cache noncacheable
> >
> > Which will ignore cache-control, Expires: immediately, and the like.
> > Or, is it a HTTP protocol violation?
>
> Taking into account the vast number of webmasters who have to present
> each month a lot of hits to the advertisment or marketing droids, we'll
> probably see more and more servers with "vanity Expires".
>
> So, such an option in Squid, is, IMHO, necessary. It has to be tunable in
> order to protect the legitimate interests of normal webmasters, which use
> Expires for legitimate reasons.

Agreed. I'm working on the new software for Cybank, ATM. Unfortunately,
someone before my time made a design decision that the response to a
given URL is built 'on the fly', and carries no-cache, and Expires:
quite-some-time-ago headers. I _have_ managed to convince the
administration to put a LOCATION: header field into the responses with a
rotating 32 bit counter. This should make it reasonably cache-friendly
(especially as the average document size is around 30 bytes, plus
headers).

I am however open to better ways to do this that don't involve locking
horns with the boss on the fact that a single URL produces a
multiplicity of responses (albiet very small ones). I'd lose that
battle, but if I can arrange some other form of behaviour to make the
system cache-friendly, and not attracting of the ire of sysops, I'll
gladly do so.

D
Received on Tue Oct 29 1996 - 04:35:18 MST