- From: Roy T. Fielding <fielding@kiwi.ics.uci.edu>
- Date: Tue, 10 Mar 1998 14:43:19 -0800
- To: Jeffrey Mogul <mogul@pa.dec.com>
- Cc: http-wg@cuckoo.hpl.hp.com
It is impossible to tell right now whether failing to address the issue will leave several contradictions in the next draft, or if those contradictions will be fixed by other changes. Either way, there is only one order of precedence (and thus implementation) which will result in a conservative and efficient test of conditional requirements given the presence of both last-modified and etag. Spending effort to not specify that in the draft is a bit silly, but won't change the correct implemenations one iota. Specifying the interaction between mutually contradictory conditionals (If-Match/If-None-Match, If-Modified-Since/If-Unmodified-Since) does not describe a plausible scenario, but it is just as easy to say that the result of combining such conditionals is explicitly undefined, rather than saying that there is no precedence (which says nothing). Specifying the interaction between etag-based conditionals and last-modified-based conditionals is plausible (and indeed specified elsewhere in the draft, just not where the reader can easily find it) since such combinations will be the norm when using a proxy. ....Roy
Received on Tuesday, 10 March 1998 14:53:30 UTC