RE: squid cache retrieval in place of errors

On Friday, 11 October, 1996 10:09, Christian Balzer[SMTP:cb@brewhq.swb.de]
wrote:
> An option like this would be nice (and has be requested just recently
> before), though it would be even nice if pages such retrieved could be
> prepended with some message (heading) stating that this is a stale
> copy and not the real thing...

There is a very good paper which was presented at the Fifth International
World Wide Web Conference which describes a system developed by AT&T based
on Harvest cache which will indicate which links on a page are available in
the cache and which would require external connection.

Squid will currently give you stale data (without warning), if ttl hasn't
expired, and has been stated it is generally better to get old data rather
than nothing at all, especially if the old information is available in the
cache. If a user really wants immediate fresh information then
Pragma-No-Cache will solve their problem.

I have included the abstract but for full details have a look at:
<http://www5conf.inria.fr/fich_html/papers/P47/Overview.html>

Mark

===================================================================
TeleWeb:
Loosely Connected Access
to the World Wide Web

Bill N. Schilit, Fred Douglis, David M. Kristol,
Paul Krzyzanowski, James Sienicki, John A. Trotter

AT&T Bell Laboratories
600 Mountain Avenue
Murray Hill, NJ 07974

Abstract: The development of the World Wide Web (WWW) has made people
reliant on continuous, high-speed, low-cost networks in order to get work
done. Ideally, one should be able to browse the Web anytime, anywhere,
whether connected to such a network or not. This paper describes the design
of TeleWeb, a system we are building to support this goal. We believe that
fine-grained cost control is crucial and have developed a "reactive
architecture" that supports user-specified adaptation under various
operating conditions. There are four key features to TeleWeb's
architecture: costs are made visible to the user through annotated HTML;
budget monitoring warns the user when operations exceed pre-specified
limits; actions may be postponed and later triggered when conditions are
met; and finally, user customization and system configuration values may
automatically adapt according to the changing conditions of use. These
mechanisms work together to provide an asynchronous, email-style of
browsing in which users can work disconnected from a cache of documents, or
trade off communication cost against information needs.

---
Mark Purcell					m.purcell@navy.gov.au
Lieutenant Commander, RAN			purcell@rmcs.cranfield.ac.uk
+44 (1793) 783 739  (tel/fax)			http://www.navy.gov.au