Tuesday, November 15, 2005
Center of Excellence Proposal
à
Article in Datawarehouse.com on Semantic Technology
Roadmap à
Previous
comments on web service
execution
environment and ontology à [199] [200]
Reference Model
for Service Oriented Architectures
In reading the Draft (Nov 15th, 2005) OASIS Reference Model for Service Oriented Architectures, I am struck by the quality of the thinking and the presentation. I am also not feeling un-comfortable with the language. When compared to the W3C Semantic Web RDF/OWL standards documentation, the OASIS technical committee takes greater care in talking about what actually are the processes and elements of this architecture.
In particular the notion of web service is fully developed without loading onto the conceptual model, or a web service, the (I think false) assumptions made by some in the Semantic Web community regarding the need for precise definitions (and the resulting "inferencing").
<Quote, from the
Draft>
3.2.3.2 The limits of
description
There are well-known
theoretic limits on the effectiveness of descriptions - it is simply not
possible to specify, completely and unambiguously the precise semantics of a
service.
<End Quote>
I note that "chains of computational events" and "deductive inferencing" are often not the same thing. So the important aspects involved, in the OASIS draft,
{ visibility, interaction, effect }
lead to framework elements
{ capability, service, service description }
and
{ exchange, execution context, policy }
without conveying uncertain notions that some intelligence computer will apply some advanced and/or proprietary algorithm to create an inference about such things as policy and context. Humans are to do this. Commits, on the draft, describing an Adaptive Service Generation Conceptual Framework architecture are consistent with Human-centric Information Production (HIP) mediated by well defined and publicly accessable precision at the level of the services and the components of services.
Other comments on the draft make it clear that autonomous process are to be defined within a stable structure having solid ontology and thus little use for coin flipping pattern identification algorithms. The reduction of the mystery in machine-based interoperability is still hindered by language such as the use of the phrase “collaboration between roles” .
A CollaborativeProcessFlow
is a set of correlated Activities, Events and Decisions that represent
collaboration between roles belonging to (autonomous) business entities. Each
flow instance has the instance number and goes through Stages. Sequences are
used if the specific order of executions is required.
But the authors need to have the notion, of a role, well developed before the notion, of computational interaction involving a type of role types, can be stated. So the short hand seems acceptable.
I do not find an abundance of statements that are literally untrue, such as "machines that understand each other."
What I find is a framework for producing a high level understanding of what large scale enterprise architectures are and can be when web services are available or could become available.
In my perception of the world, SchemaLogic is still the most practical "service oriented architecture" confined strictly to services that allow people to understand communication between diverse individuals doing things other than IT.
The new OASIS draft is suggesting a B-2-B process management enterprise that is more than mere humans trying to communicate. So the draft is approaching the vision of Sandy Klausner's "CoreSystem".
If one thinks about the total intra national commodity transaction space in the world then one has a very large enterprise that is being modeled by some at US Customs. (Not many, but some.)
The B-2-B interaction exposed in this modeling is huge.
The stakes are high (for many reasons). The modeling will be done poorly. To do a good job would induce great change, and incumbency will oppose change. This is a property of a system, not the intent of individuals. This system dynamic is why people like myself, and others like the original authors of the Topic Map standard, have failed so far to bring a reasonable foundation to the developers of the institutionally supported ontology standards (in particular the W3C Semantic Web community).