RE: URLs and access issues

Caching does present a particular issue. A lot of content accessed today is produced by dynamic processes and consequently uncacheable. If there was a way to identify fragments within the content that could be cached then this situation could be improved, but we're not talking about a backwards compatible solution here. Only differenciated URLs could be compatible with URL-based Web caches, but differentiated URLs go against the One Web idea.

Redirection could work with old caches. (One uncacheable main URL redirecting to cacheable alternative URLs, which are not revealed to the user.)

Selection could work. (One uncacheable dynamic page that selects cacheable associated resources. Think <frames>. (OK, perhaps that was just too horrible to think of.))

What about upgrading key cache servers to employ Etag indexing (or something similar), and not just relying on URLs.

Then we have security issues. Delivery via SSL currently prevents cache sharing, and with more and more sites requiring security, caching opportunities are reduced.

So, as we move to a more secure, more personalizable, more dynamic and more "One Web"-friendly Web, we are going to need a massive overhaul of the way caching is done.

---Rotan

-----Original Message-----
From: Michael Kaye [mailto:michael@mxtelecom.com]
Sent: 19 August 2005 19:57
To: public-bpwg@w3.org
Subject: Re: URLs and access issues



Paul Walsh wrote:

>You don't need different URIs to represent
>different content - this is why we have adaptation.
>  
>
(further to my email)

Yes, you do.

If you start requiring that all caches across the internet MUST NOT
cache any content (as it could change at any time, depending on the time
of day, the device, anything) then the gain in bandwidth use is going to
be staggeringly huge.

Requiring that all mobileOK content is not cached is not a step forward
that I'd like to see occur.

Michael.

Received on Friday, 19 August 2005 19:17:42 UTC