Re: [ws-caf] Draft Minutes 18 July 2005

[Mark Little <mark.little@arjuna.com>]

Apologies for dropping off after 30 minutes yesterday. However, lots of 
potential AIs. Any timeline? Any chance of keeping the discussion going 
around these things?

>
>7. WS-Acid: feedback and issues.
>
>Those that affect CF first. Boxcarring, context reply.
>
>AG: Recommend look at BTP where these issues were dealt with successfully. The wished-for layer that should magic the boxcarring problem away still doesn't exist, some years on from BTP. Layering anyway complicated by cross-over of vectorizaton and context reply.  Like GP's desire to put this into WS-CF, and this means that we must look beyond WS-Acid to ensure we have dealt with the general case. GP: want to collect some more datapoints on issues raised by AG, for example in WS-A not clear whether possible to use the EPRs as referencing devices for specific services. Support for a metadata buckets: refs to service information from the WSDL description. Not clear that this is well-defined enough to allow identifying elements to be specified and used.  identity-like elements. GP will take action on researching that issue. GP, secondly: EPR hiding WS-A. Moving incrementally towards WS-A. 
>
>Discussion of checked transactions: BPEL serializable scopes could use atomic transactions. Asynchronous message passing poses problem of checking. Should we provides guidelines at our level. How can we take advantage of request correlation capablilities. There could also be a BPEL-specific approach. MC: Have agreed to adopt WS-A.However, when? GP they're in last call. Wait for that? 
>
>GP/AG: we must first of all resolve if checking (knowledge of inferiors present) has to be capable of preceding knowledge of application completion. If that is so, we need counting and identity of inferiors, and if that is so then AG point becomes relevant: maintaining the abstract EPR notion in order to cover potential lack of identity in WS-A is also posed.
>
>PF: two kinds of identifier: guaranteed unique, and app-generated -- both need to be made visible, not buried in the opacity of the addressing system. 
>
>
>  
>
Mark.

-- 
Mark Little
Chief Architect
Arjuna Technologies Ltd
www.arjuna.com