RE: [bt-spec] BTP Issue 16 : Negative reply to REDIRECT

[Peter Furniss <peter.furniss@choreology.com>]

 

 

BTP Issue 16 : Negative reply to REDIRECT

Category: minor technical
Submitter: Gordon Hamilton, AppliedTheory
Page: 23
Question: Is there a negative reply to the REDIRECT?

---

 

 I think there probably isn't.

 

 Failure to deliver at all (because the target isn't there any more) would be lower layer fault (e.g. SOAP, HTTP), and so would appear to BTP as communications failure.  If the message gets through to the BTP entity, but there isn't any information about the transaction in question, it doesn't matter (this is quite possible given the presume-abort nature of the recovery model). If the message does get through, and there is information that ought to be changed, but it can't be, then there's something wrong with the implementation (possibly only transiently).

 

we could generate a BTP general FAULT for the last case, but we don't actually know where the message has come from, and they can't do anything about it except resend it. REDIRECT is basically a one-way message.

 

Peter