RE: ISSUE: Where do faults go?

Hi Chris:

> > > IMO, if a SOAP message contains WS-A headers that are 
> inconsistent 
> > > with the spec in any way that the generated fault SHOULD 
> be sent to 
> > > the endpoint identified by the relevant SOAP MEP/binding.
> > 
> > This I'm not sure I agree with.  If I send you a single, valid, 
> > non-anonymous FaultTo header, and duplicate ReplyTo headers (or any 
> > other WSA screwup), wouldn't you want the fault on the 
> FaultTo EPR?  I
> 
> Possibly, but what if the fault is generated *before* the 
> [fault to] is parsed/processed? Are you suggesting that if a 
> fault is generated that processing of the WS-A headers is to 
> continue? That seems a little odd to me. If you have an 
> endpoint that is sending garbage in regards to [reply 
> endpoint], what expectation would you have that the remaining 
> MAPs are coherent? I'd just as soon consider the whole lot as 
> tainted and send the fault as if WS-A were not being used. 

Well, I'm pretty sure that we don't say anything remotely resembling
this.  If we wanted this behavior (ignore ALL WSA headers if there's a
fault with ANY of them) we would need to make this "WSA processing
model" explicit.  I'm not saying it would be a bad idea, even - just
that it would be a serious change.

> > suppose actually you need a valid FaultTo and also a valid 
> MessageID 
> > for correlation in that case, though - but assuming those, wouldn't 
> > you send other faults to FaultTo?
> 
> I'm not convinced. However, the case I was making, possibly 
> poorly, is that no matter what [fault to] says, that the 
> sender of a (request) message MUST be prepared to receive a 
> fault as if WS-A were not in use. 

That is certainly true, though orthogonal to the point above, I think.

--Glen

Received on Monday, 13 March 2006 21:11:24 UTC