- From: Cullen Jennings <fluffy@cisco.com>
- Date: Wed, 16 Nov 2005 12:37:34 -0800
- To: Julian Reschke <julian.reschke@gmx.de>, Lisa Dusseault <lisa@osafoundation.org>
- CC: Geoffrey M Clemm <geoffrey.clemm@us.ibm.com>, WebDav <w3c-dist-auth@w3.org>
On the call this morning, I thought that Geoff outlined 3 proposals 1) change nothing (assuming says MUST put in header) 2) don't change how anything works today but add the clarifying statement that servers MAY NOT put it in a body 3) change to say that servers MUST put in both body and header Now I might have the descriptions of these three all wrong but I am pretty sure that I heard that 1 or 2 worked for someone and 2 or 3 worked for someone else and that 2 worked for everyone. I thought we decided to put 2 in the next rev of the *draft* and then ask the WG (folks on the list including people on this thread) had a problem with that draft. What part of this did I not get? What are we still discussing? On 11/16/05 12:23 PM, "Julian Reschke" <julian.reschke@gmx.de> wrote: > > Lisa Dusseault wrote: >> I checked back in the email archives, and it looks like there was a very >> closely related discussion which I thought covered it but it didn't get >> into this specific issue. Naturally consensus needs to be found in the >> WG and I'm sure we can accomplish this. So are there any other opinions > > (nod) > >> on whether the server MUST return the lock token in the body as well as >> the header of a response when a new lock is created -- even though we >> all agree the server doesn't need to normally put that info in the >> lockdiscovery property? > > Making a normative change to the spec because of specific broken clients > IMHO is as bad as the other way round. Just because we say so will not > resolve the interop issue, if it still exists. Does anybody know whether > this is still the case? > > To me, this seems to a a perfect example, for an interop guide, not a > change to the spec. > > Such as: > > 1) What to do if the LOCK request suceeds, but no Lock-Token respnse > header is there (answer: complain to the server vendor, and then in the > worst case check the response body). > > 2) What to do if a PROPFIND response contains *multipple* > <lockdiscovery> properties? (answer: complain to the server vendor, > and/or assume that this is this server's idea how to return info about > shared locks). > > We've talked about a document like that several times. Should I start one? > > Best regards, Julian
Received on Wednesday, 16 November 2005 20:37:44 UTC