Re: Ideas/solutions to caching identical objects ...

From: Serge A. Krashakov <sakr@dont-contact.us>
Date: Mon, 2 Dec 1996 10:33:19 +0300

George Michaelson <ggm@connect.com.au> wrote:

> I raised this question on the nlanr cache list a few months back. The answer
> then was to wait for the MD5 hash signing code to be stable, and then use
> this to verify documents in transit by some logic comparing URL, MD5 hash
> and some level of document meta-information. (for MD5 read whatever hash
> is decided on, hopefully cheap to compute)
>
> You need the hash, and in-doc meta info since the URL may vary across
> different mirror instances. Once there is something like a URN inside a
> document then you have a version-specific field to compare with a hash
> function and a URL.
>
> Actually, I guess you could use the URN if it was trustably globally
> unique but a hash seems safer.

I think that proper cache server must keep also such information as
"Contenet-Language" and "Content-Type: ...; charset=" fields because many
multilanguage servers return documents in different languages/charsets
with the same URL.

Serge Krashakov
Received on Sun Dec 01 1996 - 23:57:05 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:33:48 MST