[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [ebsoa] Scope of TC
2 positive comments in line. Duane writes: "This is a good conversation however I think we had agreed in New Orleans that this TC will use the patterns syntax to define our first work, not formal specifications of other groups. There were several really good reasons for this and it does not limit us in any way. Let me explain. 1. We keep everything defined using a non technology specific pattern. Example: abstract = sending a message The implementors can use these to implement a concrete messaging format. concrete = ( ebXML Messaging || SOAP ) The item they need to correlate is the feature sets and their requirements (something BCM can help with)." +1 I had not read this connection between patterns and abstraction level before. Sounds like a good approach to me to not tie architecture component/functional specification to implementation. The last sentence is a little too cryptic for me. Will BCM define the feature sets and requirements for ebSOA? Duane writes: "2. The danger is we rely on other specifications is that we are no longer a valid architecture if they change their normative functionality outside the scope of the architectures requirements. We are defining patterns and blueprints for what a component must do. Example: The patterns and blueprints show how a human actor sets a token into a message for a bilateral service level agreement. SOAP does not have this function, therefore we have a dilemma. ebXML messaging does and is a good fit for use within our architecture although it is not normatively called out." Your main point about getting involved in relying on specifications sounds OK, but the example is possibly misleading. I think that WS arena has several candidate technologies for bilateral service agreements. WSPL, XACML and eventually WS-Policy may partly implement that functionality. The W3C announced an October discussion meeting on these topics, so there may be an(other) open specification on these matters. [I know there is interest in the ebXML CPPA as to how one could xslt transform between the approaches if it becomes just a matter of tag names and syntax.] WS-CDL may also partly implement process agreements. Maybe WSDL, being an interface/binding definition language, can be viewed as specifying a contract for which a human could set a token (endpoint value, e.g.). ebXML has CPPA and BPSS to specify aspects of the contract. Initially I think the totality of WS specifications seemed to be missing a few parts present in the collection of ebXML specifications, but gradually the collections are moving toward a functional parity. ebXML's current convergence tactic (3.0) is to adopt functionally equivalent WS specifications to reduce the implementation burdens on developers and users alike. I find it particularly ironic as the two sets of specifications grow functionally comparable, that architectural people are urging the ebXML stuff to be thrown out. On what basis? I would hate to see ebSOA degenerate into the kind of technical marketing buyer's guide that the w3c architecture document did (captured perfectly in their mantra that a web service is something that uses SOAP and WSDL and UDDI, or at least 2 out of 3).
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]