I had opportunity to think about this subject in early 2004, during the student project work. The question is very important and is responsible for the core direction during the implementation.
In contrast to WS-AT providing ACID properties, where the behavior of the coordinator and participant is prescribed by the protocol and the only question is how to finish the 2PC protocol on the participant side (according to the native operations on the resource) it is not that easy for the WS-BA. Further more the outcome of a BA can be different to each participant pair due to the relaxation of ACID. In general this outcome is dependant on the application semantic and therefore the decision can not be met by a "semantic-less" coordinator. In the solution proposed in (http://jwst.sourceforge.net/resources/publications/acm-wsba-paper.pdf) the responsibility of each decision is delegated to the initiator by a usage of an interposed middleware component. In my opinion this is a general case, and a special case could be a set of predefined strategies for the coordinator. Another possibility to solve this issue is to let the coordinator "understand" the semantic of the activity coordinated by reading a kind of a process description (e.g. BPEL). Kind regards, -- Dipl. Ing. Simon Zambrovski Research Assistant Hamburg University of Technology Institute of Telematics Schwarzenbergstrasse 95 D-21073 Hamburg Phone: +49 40 428783704 --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
