RE: operation name in responses?

Hi Jacek!

> thanks for the reply. I somehow didn't think of the 
> correlation that was necessary (but I'm kinda unsure here - 
> is it really necessary?) in which case it takes care of the 
> rest of the messages regarding the operation name.

Yup - I think that the MEPs have to imply correlation, otherwise they
don't really work (without correlation you're done after the first
message with not much hope of relating faults/replies to your active
operation).  Should we be saying more about this in the spec, you think?
The *method* of correlation (transport semantics, SOAP headers, etc) is
up to you, but the correlation is built in to the concept of MEP....

--Glen

> On Mon, 2004-02-23 at 17:06, Glen Daniels wrote:
> > Hi Jacek: 
> > 
> > > I just thought - for the convenience of the service, we 
> require that 
> > > every operation carry the operation name on the wire (or 
> other means 
> > > of identifying the operation name). Does this apply to 
> the response 
> > > messages of request/response MEPs (the out message of 
> in/out, that 
> > > is); or in fact for all messages in any message exchange 
> pattern? I 
> > > think it should.
> > > 
> > > Whichever way it is, is the spec (or proposal, I don't 
> know where we 
> > > stand now) clear on this point?
> > 
> > It's a proposal at present, and no we don't specify that 
> the name must 
> > be carried on any messages except the initial one.  I think 
> that's the 
> > important one, and further messages which might correlate 
> to that one 
> > already have implicit needs for somehow making the 
> correlation work...
> > once that's done the operation is obvious.  We do need to 
> make clear, 
> > I think, that it's the initial message which is important, whether 
> > that be inbound or an outbound.
> > 
> > --Glen
> 
> 
>

Received on Monday, 23 February 2004 11:22:54 UTC