- From: Kohei Honda <kohei@dcs.qmul.ac.uk>
- Date: Mon, 30 Oct 2006 16:45:38 +0000
- To: Yves Lafon <ylafon@w3.org>
- CC: Martin Chapman <martin.chapman@oracle.com>, 'Gary Brown' <gary@pi4tech.com>, 'WS-Choreography List' <public-ws-chor@w3.org>
My apologies I came back to this two weeks later. I just realised these mails. I do not know the initial motivation for this flag. Anyway here I note how we can use it. I am sorry I am presenting it now, but I am getting some ideas about the use and theory of CDL better than before, and that is giving me additional ideas. Having initial flag combined with the (automatically filled) primary session identity is in fact very useful. That gives a flexible mechanism to extract a series of interactions from a given description and further associate a clear-cut idea of type. Suppose Alice asks Bob three times, each with initiation flag. Suppose, to the same channel, Carol asks Bob twice, each with initiation flag. If these interactions are all (say) askQuote followed by thisIsQuote, then this means Bob does not have to offer "two repeated interactions" and "three repeated interactions" separately to Alice and Carol. It only offers one kind of interaction, repeatedly offered to anybody who invokes Bob. If this is totally nested, we can use perform. But suppose two of these interactions (especially assuming each is in fact containing a series of many communications) interleave with each other. Then we cannot use perform. Because of loops and multiple participants in loops, this can become arbitrarily complex. It can be disentangled if we are clear about different sessions --- where each begin, to which session each message belongs to, etc. *Practically* this means the following: If we have this flag, and if we have co-relation identity (as is called now), then we can have a very clean type structure and EPP mapping. If this is harmless otherwise, therefore, I propose to keep it. kohei Yves Lafon wrote: > > On Tue, 17 Oct 2006, Martin Chapman wrote: > >> >> I very much agree with the reuse argument to remove the flag; a >> choreo will >> be far more resusable if the flag doesn't exist, or at least add a >> don’t'care/maybe option! > > I concur, In fact, the best, if this distinction is needed would be an > initiate element somewhere (like on a description of the root choreo) > with a link to the initiating interaction. But not something inside > the choreo definition. > >> >> Martin. >> >>> -----Original Message----- >>> From: public-ws-chor-request@w3.org >>> [mailto:public-ws-chor-request@w3.org] On Behalf Of Gary Brown >>> Sent: Tuesday, October 17, 2006 9:04 PM >>> To: 'WS-Choreography List' >>> Subject: 'initiate' flag issue >>> >>> >>> >>> Hi >>> >>> Thanks to Yves, I have found Nick's comments on my original >>> proposal for >>> removing the initiate flag: >>> http://lists.w3.org/Archives/Public/public-ws-chor/2004Oct/0049.html >>> >>> Nick's argument focused on the declarative flag helping a >>> designer when >>> composing choreographies. However, I don't see this as being a >>> compelling argument - in fact, it is better from a reuse point of view >>> that previously standalone choreographies that initiated a >>> choreography >>> session are able to be composed into more comprehensive choreography >>> which now starts the session at a different point. >>> >>> Regards >>> Gary >>> >>> >>> >>> >> >> >> >
Received on Monday, 30 October 2006 16:45:58 UTC