- From: Burdett, David <david.burdett@commerceone.com>
- Date: Fri, 16 May 2003 11:19:52 -0700
- To: "'Ricky Ho'" <riho@cisco.com>, "Burdett, David" <david.burdett@commerceone.com>
- Cc: public-ws-chor@w3.org
Ricky You said ... >>>Or do we define 1 choreography (for A/B/C) and then apply a filter to extract another choreography for C/B ?<<< Although you could do this, I am wondering what the benefit would be? For example there might be several different ways in which C&B could exchange messages to realize the same end goal. "A" should not know about this unless there is a real need - i.e. it affects, in some way, A's behavior. If A's behavior is not affected, then it means that any change to the B/C choreography should result in A being informed since he has the complete A/B/C choreography. If on the other hand A does need to know, then you can might want to think of the A/B/C choreography as an *extension* of the B/C choreography. I always think that extension is better than subsetting or filtering as it promotes reuse and is, I think, actually easier to do. David -----Original Message----- From: Ricky Ho [mailto:riho@cisco.com] Sent: Thursday, May 15, 2003 9:04 AM To: Burdett, David Cc: public-ws-chor@w3.org Subject: RE: pi-calculus ... Lets say there are 3 roles A, B, C in a choreography. A and B need to know the detail of every other role but C only need to know about B. Do we need to define 2 choreographies (one for A/B/C, and the other for C/B) and verify their conformance ? If the answer is "yes", then I agree that this is a question of good design practice. Otherwise ... Or do we define 1 choreography (for A/B/C) and then apply a filter to extract another choreography for C/B ? If the answer is "yes", then we need to support the "segmentation" concept in our choreography model. Then Jon's concern is very valid. Rgds, Ricky At 09:54 AM 5/13/2003 -0700, Burdett, David wrote: >Jon > >I agree. I think that publication SHOULD be on a "need to know" basis. If >your internal business process involves carrying out choreographies with >roles that other roles have no need to know about then they should be >defined as separate choreographies that are under the overall control of >your internal business process. > >Really though I think this is primarily a question of good choreography >design rather than a constraint on what our choreography spec should allow. > >David > >-----Original Message----- >From: Jon Dart [mailto:jdart@tibco.com] >Sent: Tuesday, May 13, 2003 9:45 AM >To: Assaf Arkin >Cc: Burdett David; 'Jean-Jacques Dubray'; 'Nickolas Kavantzas'; 'Steve >Ross-Talbot'; public-ws-chor@w3.org >Subject: Re: pi-calculus ... > > >The only caveat I'd make here is related to the earlier discussion about >visibility. In a B2B scenario, I may not want to publish to all >participants a definition that includes "all the behaviors of all the >roles". > >--Jon > >Assaf Arkin wrote: > > > > Burdett, David wrote: > > > >> So how about as a requirement ... > >> > >> "The choreography specification must provide a method of defining the > >> rules and sequence for exchanging messages between two or more roles > >> (e.g. buyer, seller, etc) that: > >> > >> a) Provides a single choreography definition that covers all the > >> behaviors of all of the roles > >> b) Treats each role equally > >> c) Specifies messages in an abstract way, i.e. independent of the > >> precise data structures and transport binding." > >> > >> David > >> > > Of course I would say +1 to this, since it defines the problem we're > > trying to solve and provides a good pattern for any solution. > > > > arkin > > > > > > > >
Received on Friday, 16 May 2003 14:19:54 UTC