- From: Miles Sabin <miles@milessabin.com>
- Date: Wed, 17 Jul 2002 14:52:19 +0100
- To: www-ws-arch@w3.org
This seems to relate to quite a few issues that've been discussed here recently (late vs. early binding, loose vs. weak coupling, and equivalence) so forgive me if this is a bit vague. It's noticeable that as we move up through the layers of XML/WS specifications from typing of individual messages, through specifications of individual multi-message operations, to specifications of multi-operation conversations, that the precise types of items from lower layers are being hardwired into the higher layers. For example, a WSDL port-type/operation hardwires in the types of its input and output messages; and in all the orchestration languages I've seen so far, specific WSDL port-type/operations are hardwired into the specification of the conversation. In the latter case, in particular, this seems to be particularly unfortunate, given that many business processes will be very similar at the level of orchestration, differing only in the types of the concrete service endpoints they apply to (which in turn might well only differ in message type) ... binding those lower layer types into the definition of the higher level orchestration makes it impossible to reuse the orchestration in different contexts. If this is the case then it might be desirable to take a leaf from the generic programming community's book, and consider defining higher layers relative to type-parameters rather than concrete types. WRT orchestration languages this would allow us to specify generic conversations schematically, effectively abstracting over broad classes of business processes, then instantiate at particular concrete port-types/operations to get a concrete conversation. Thoughts? Cheers, Miles
Received on Wednesday, 17 July 2002 09:52:51 UTC