- From: Tommy Lindberg <tommy.lindberg@gmail.com>
- Date: Mon, 17 Jan 2005 11:27:52 +0000
- To: Berin Lautenbach <berin@wingsofhermes.org>
- Cc: "www-xkms@w3.org" <www-xkms@w3.org>
Hi Berin - It was I who sent you those StatusRequest's. I meant to send you the message exchanges this morning but now I guess I don't have to. > StatusRequest processing is fairly ill-defined in the spec. Are you sure you have the most recent version of the spec? Guillermo raised an issue some time ago that removed a discrepancy in this area of the spec. I don't know if this answers your question(s) but this is how I think about it: The StatusRequestType extends PendingRequestType leaving ResponseId as a required attribute in the former. Consequently, ResponseId should always be part of your lookup criteria. There is no ResponseId attribute in any other request type than the two mentioned above, so the OriginalRequestId is the only option for tying together the whole sequence of messages in a combined two-phase/asynchronous exchange. > and I just tracked the problem down. I'll try again later. Regards Tommy On Mon, 17 Jan 2005 19:10:53 +1100, Berin Lautenbach <berin@wingsofhermes.org> wrote: > > Peoples, > > I've had my somewhat flaky server hanging on me fairly regularly, and I > just tracked the problem down. > > StatusRequest processing is fairly ill-defined in the spec. I > originally assumed that the OriginalRequestId attribute should be used > to lookup the original request. Someone has been sending me a > StatusRequest without the OriginalRequestId set, but with ResponseId > set. My server wasn't properly checking that everything it needed was > there, so it got unhappy. I have now updated to also lookup based on > ResponseId if the OriginalRequestId is not there. > > However I'm curious as to which is correct, particularly given > interactions between two-phase and asynchronous. I.e. OriginalRequestId > is set to InitialRequest/Phase 1 for everything later - which confuses > any async lookups based on OriginalRequestId. In which case - if the > semantics of OriginalRequestId change depending on mix of two phase and > async, why not use OriginalRequestId for one and REsponseId for the other? > > Does all of that make sense, or have I completely missed the point > somewhere along the line? > > Cheers, > Berin > >
Received on Monday, 17 January 2005 11:28:24 UTC