- From: Assaf Arkin <arkin@intalio.com>
- Date: Fri, 20 Dec 2002 20:16:11 -0800
- To: "bhaugen" <linkage@interaccess.com>, <www-ws-arch@w3.org>
> I think there's a critical issue buried in this dialog: > how many parties need to agree on how much detail > of a particular set of coordinated interactions? > > It's like, how many parties need to agree on this contract? > Or, some variation on a need-to-know rule? > > Or, to put it differently, consider a range of > configuration options from: > 0. zero-config (something like Mark's models), to > 2. pairwise external interaction agreements with > coordination as internal responsibilities, to > M. many-parties-agreeing-on-all-details-of-all-interactions > (is that the BPEL model?). > > Which configuration options would be suited > to which situations? Good question. From what I've seen so far, there are four different use cases. If your process is ad hoc or you don't automate it, then you don't ask anyone to agree with you (zero-config). If you provide a service to a large audience then you will typically have one party deciding on the interaction agreement. See Amazon, eBay, etc. When two (or even more) partners are working together through an interaction that is specific to them you will typically see two-way coordination. Even if you have three parties, it's easier to start by having each pair agree to something and then let the other two stitch it. Not a prefect approach, just practical. Some scenarios are defined by consortia and them get adopted by businesses. For example RoessetaNet or supply chain management. In this case you have a multi-party definition and each partner respects their role and don't try to break it by having a more specific interaction that is different form what every other partner (actual or possible) would expect. arkin
Received on Friday, 20 December 2002 23:17:31 UTC