- From: Paul Burchard <burchard@cs.princeton.edu>
- Date: Wed, 29 May 1996 04:33:59 -0400
- To: masinter@parc.xerox.com
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Objection: Methods which are not idempotent, at least in principle, must not be automatically retried by client software without human intervention. (See my earlier post bringing up this objection: <URL:http://www.ics.uci.edu/pub/ietf/http/hypermail/1996q2/0336.html>; Koen Holtman suggested specific changes (too strong) in: <URL:http://www.ics.uci.edu/pub/ietf/http/hypermail/1996q2/0357.html>. I don't see how the current draft addresses this objection.) Proposal: Section 8.1.5 (para. 4), and Section 8.2 (initial paras.) need to be amended to specify that the automatic retry requirement applies ONLY to methods considered idempotent; other methods MUST NOT be automatically retried, although clients MAY offer a human operator the choice of retrying the request. (The end of Section 9.4 would not then need to be changed.) Section 9.2.1 should identify both "safe" and "idempotent" as distinguished semantic properties of methods, and define GET, HEAD, PUT, and DELETE to be "idempotent" in the sense that (aside from error and expiration issues) the user's responsibility for N>0 identical requests is the same as that for 1 request. -- --------------------------------------------------------------------- Paul Burchard <burchard@cs.princeton.edu> ``I'm still learning how to count backwards from infinity...'' ---------------------------------------------------------------------
Received on Wednesday, 29 May 1996 01:39:23 UTC