- From: Julian Reschke <julian.reschke@gmx.de>
- Date: Fri, 25 Feb 2005 21:34:57 +0100
- To: Mark Nottingham <mnot@mnot.net>
- CC: w3c-dist-auth@w3.org
Mark Nottingham wrote: > > A couple of questions about WebDAV; > > * The only mention of idempotency in any WebDAV RFC that I see is: > "Responses from a MKCOL request MUST NOT be cached as MKCOL has > non-idempotent semantics" in 2518, Section 8.3.2. Is it the case, then, > that all other WebDAV-defined methods are in fact idempotent? > > I could see MOVE as non-idempotent, IF the semantics of COPY are such > that it's legal to COPY a null resource; i.e., the MOVE would first COPY > the existant resource, then DELETE it, then on a subsequent request, > would COPY the null resource. However, it isn't clear from a casual > reading of 2518 as to whether COPY will fail (404?) if you try to point > it at a null resource. > > * How is MKCOL non-idempotent? I.e., how does performing MKCOL /foo five > times have different side effects than performing it once? 2518 says > that "If the resource identified by the Request-URI is non-null then the > MKCOL MUST fail." It seems that all MKCOLs after the first one will > fail, thus bringing no new side-effects. I can certainly see how MKCOL > could be involved in a non-idempotent sequence, but don't see how it is > on its own; am I missing something, or should it be "MKCOL has side > effects" instead of "MKCOL has non-idempotent semantics?" You are right, MKCOL isn't idempotent, nor is MOVE. I'll raise issues against RFC2518bis, RFC3253bis, RFC3648bis and RFC3744bis so this get's fixed in future revisions. In BIND (<http://greenbytes.de/tech/webdav/draft-ietf-webdav-bind-11.html>) we're hopefully already doing it right; but you may want to check :-) Best regards, Julian -- <green/>bytes GmbH -- http://www.greenbytes.de -- tel:+492512807760
Received on Friday, 25 February 2005 20:35:37 UTC