- From: Kohei Honda <kohei@dcs.qmul.ac.uk>
- Date: Mon, 30 Oct 2006 20:30:38 +0000
- To: Gary Brown <gary@pi4tech.com>
- CC: "Monica J. Martin" <Monica.Martin@Sun.COM>, Martin Chapman <martin.chapman@oracle.com>, 'WS-Choreography List' <public-ws-chor@w3.org>
Hi Gary, Well that I feared. Co-relation identity is not part of the type structure but part of application data. I think this cannot be changed now. I am sorry I could not predict this at the initial point. Anyway here is a concrete situation where such identity (at least a basic one) is better be as part of type structure, part of a header, something users cannot manipulate. Suppose we send a channel (which means we delegate a session to somebody in practice). Since the identity is just part of a message content, we do not have an automatic scheme to send this identity together. We do not have to do it. So it is not ensured that identity is preserved --- if we wish to do it. Of course we can always do this in an ad hoc manner, but the issue is we cannot guarantee the identity of a series of conversations. So generally, at least some part of identities should be part of "message header", something which users cannot manipulate and part of the interaction structure (hence that which can be type abstracted). No other web languages have it, so I think we should leave this as it is. As I wrote, my point is that this is the general way to treat such protocols, *not* we should do this for this particular instance. The generality is that we can deal with arbitrarily inserted actions between request and return, and still can keep the correspondence. So this would be for future. Best wishes, kohei Gary Brown wrote: > Hi Kohei > > I don't think co-relation can be used as this would be dependent upon > the contents of the messages, which cannot be relied upon to be > distinct from a previous request. So I think the most appropriate > approach is the new exchange action type, which makes it then > consistent with the way other Message Exchange Patterns are described > (i.e. explicitly). > > Regards > Gary > > > Kohei Honda wrote: >> >> A bit confused with negation: >> >>> >>> From this viewpoint, my question is: are there any these >>> "request-reply" and its variants, including >>> notifications, which cannot be captured as a pattern of interaction, >>> which can be made explicit by the >>> use of co-relation identity? >> >> I meant, in the last clause: ..., which cannot be made explicit by >> the use of co-relation identity? >> >> My question was, therefore: whether all can be captured by >> co-relation (or session) identities or not. >> As written, even if all can, I do not oppose having explicit >> constructs for specifying local (or micro) >> protocols. >> >> kohei >> >> >> >> >> >
Received on Monday, 30 October 2006 20:30:51 UTC