- From: Mike Gahan <ccaamrg@ucl.ac.uk>
- Date: Fri, 13 Dec 1996 11:53:01 +0000
- To: www-talk@w3.org
Brian Morin wrote:
>
> Peter Wrote:
> "Isn't mirroring just a special case of caching?"
>
> No, two major differences I can think of (between mirroring and proxy caching.)
>
> 1) Mirrors do not have to worry about keeping data current.
>
> Assuming that updates are sent from the primary site, there no need to querry the primary site to determine if data has expired. Given that most web transactions involve small files, the cost of opening a TCP connection and finding out if data is still current is not trivial.
>
The principal difference is from the point of view of the browser. A
mirror must be explicitly selected, outside of the normal caching
mechanism. If the mirror is viewed as a special purpose, limited topic
cache, it could be accessed automatically by existing auto-proxy
configuration mechanisms. The trick here is that there need to be
centrally maintained auto proxy config files which know about all the
topical caches (mirrors) in their network locality. Browsers oin that
locality then download the config, and, hey presto, they access the
optimal mirror automatically.
How topical caches maintain currency is a matter entirely between them
and the primary. I expect that the arrangements will be more formal
than for general purpose caches, and appropriate expiry headers
can be transmitted which reflect the general stability (or otherwise)
of the data.
--
Mike Gahan
Information Systems Division
University College London
http://www.ucl.ac.uk/~ccaamrg/
Received on Friday, 13 December 1996 06:53:33 UTC