- From: Stanislaw Ambroszkiewicz <sambrosz@ipipan.waw.pl>
- Date: Tue, 29 Apr 2003 13:53:20 +0200 (CEST)
- To: public-ws-chor@w3.org
- Cc: sambrosz@ns.ipipan.waw.pl
> To repeat myself, I think we need a really, really Good Reason to > re-introduce "service type", ... Perhaps you will not find "really Good Reasons" as long as you follow the imperative approach to service composition, like in WSCI and BPEL. In the imperative approach the task (goal) realized by a composite service need not be explicitly stated; it is hard-coded in a language that models a process in a more or less centralized way. This approach may be characterized by the phrase: "How to realize?" There is also declarative approach to service composition followed by SWORD and partially by DAML-S. It may be characterized by "What to realize?" Task must be explicitly stated in a declarative way like in Prolog. Here, the notion of "service type" is necessary in its full generic meaning to express for example precondition and post condition (effect) of service invocation. The concept of precondition and effect (see DAML-S and SWORD) is appealing. Precondition describes input whereas effect describes output. They are suppoesd to be formulas in a language, e.g., first order logic. These concepts can be also used for realizing query, i.e., a service may be asked if it can realize some condition as its effect, i.e., if this condition follows from one of its possible effects. If there is such effect then the service may reply with the corresponding precondition, i.e., the constrains on input necessary to realize the required effect. Best regards, Stanislaw -- Stanislaw Ambroszkiewicz Institute of Computer Science, mailto:sambrosz@ipipan.waw.pl Polish Academy of Sciences http://www.ipipan.waw.pl/mas/
Received on Tuesday, 29 April 2003 07:57:55 UTC