- From: Koen Holtman <koen@win.tue.nl>
- Date: Thu, 23 Dec 1999 13:44:12 +0100 (MET)
- To: "Josh Cohen (Exchange)" <joshco@exchange.microsoft.com>
- Cc: fielding@kiwi.ICS.UCI.EDU, joshco@exchange.microsoft.com, http-wg@hplb.hpl.hp.com
Josh Cohen: > [...] >I guess if there is any chance that the response could be different, >based on client auth, client type, or whatever, then it is not safe to >cache. It _is_ safe to make the response cacheable as long as you use Vary correctly, this is what Vary was invented for. >(especially since caches dont filter based on accept before >returning responses) As Roy said, most (all?) 1.1 caches don't actually implement the refined filtering made possible by Vary. They implement the Vary requirements in 1.1 by treating 'Vary: anything' as equivalent to 'no-cache'. But you can still use Vary if you want to help possible future caches: I would consider this to be good protocol design. You can find some examples of the use of Vary+Expires in RFC2295. Koen.
Received on Thursday, 23 December 1999 04:52:11 UTC