- From: Steven Ball <steve@cs.anu.edu.au>
- Date: Sat, 23 Nov 1996 15:23:22 +1100
- To: dank@idirect.com (Dan Kolis)
- cc: www-talk@w3.org, steve@ricetub.anu.edu.au
Dan Kolis writes: > I'm looking at HTTP1.1 HTTP/NG andam really interested in CATV style 100K > seat multicast and non multicast. > I keep looking for hooks in your spec for HTML so the following is totally > reliable. Very often a HTML author will reuse lots of components, graphics; > (typical JPG GIF), etc. Now Each uses its own GET. > [...] I've just finished reading the HTTP/1.1 spec, and what you're asking for can be done in HTTP without resorting to HTML hooks. You should read up on HTTP/1.1's caching design (Chap. 13, p. 50) and the various headers that affect validation and freshness (such as ETag [14.20, p. 79], etc.). HTTP/1.1 specifies persistent connections, so by default each document is fetched across the same transport connection (or perhaps across two connections for concurrency benefits). You claim you'd like to be able to "steal" documents from another site but include their ID tag so that a browser will know that it already has the document cached. However you also claim that you'd like it to be reliable and these are contradictory requirements. What if two origin servers use the same ID for completely different documents? In short, HTTP's caching model does what you want, but in general not for documents replicated across servers. Of course, if two or more servers were under the same administrative domain and the browser knew this then an application-specific solution could be created. Cheers, Steve Ball, Dept. Computer Science, Australian National University
Received on Friday, 22 November 1996 23:23:33 UTC