Alternative to the TBL RDF layer cake model of the Internet à
Part of a US Federal CIO Council discussion
(Extending this bead
thread after a period of seven months)
Previous posting was
May 12th 2006 à [57]
Community centric
service methodology à [home]
Part of an ongoing US Federal CIO Council discussion
To the SOA CoP e-forum.
Brad's comments and Joshua's comments both point at the class subclass assertion as having some technical difficulty. The omission of the discovery aspects of services may be due to an non-removeable characteristic of OWL, and this characteristic due to the foundational, and often overlooked, assertion of a specific type of class-subclass ontology.
Let us take the actor, service, effect ontology (page 23 in Chirs Harding’s recent presentation:
http://www.opengroup.org/projects/soa-ontology/doc.tpl?gdid=12153
In each case we need a mechanism for formalising "information about actor", "information about service" and "information about effects". As Chris Harding has pointed out in his presentation; we are modeling information about information.
This brings us to the difference between the topic map stand and the W3C ontology standards.
The topic map standard starts out with the distinction that an actor may be some human person external to the computer and to language. As Wittgenstein was fond of pointing out, language is used to point to real world things, and so a precise correspondence between symbols and real world things cannot be always expected. Some times one may get utility from making this expectation of precise correspondence, but the ontological assertion is not proper - due to the nature of the real world.
The foundations of mathematics and logic seem to demonstrate this distinction between the concept and the reality in several ways.
The concept has a lower or a higher degree of utility depending on many things like fidelity between represented concepts and on going reality, and completeness.
For example, the intended effect of a service may be to change the nature of an actor, so that the actor shifts its categorical placement. This is a non-locality phenomenon and occurs all the time in real social interaction.
In order to achieve this intention the actor needs transparency not only about the ontological model described, using OWL, [1] effect; but all aspects of natural reality that may be entailed to the specific use of a serves in an specific context at a specific location in a specific time.
In the social world, we find that these shifts are not following anything like the conservation laws in physics, and become non-rational in the sense of not following W3C supported description logic.
In my work, I have argued (with others) that "measurement" is the missing key not present in any OWL ontology, and for database design.
Any database design or any OWL constructions already has under gone the inductive process in the past and is “CLOSED” with respect to real time modification by the end user.
The data modeler or the knowledge engineer makes a formal construction. The motivation for this construction can and does reflect the data modeler and/or the knowledge engineer’s professional and private interests. Thus we have a problem with the origin of the design of communicative symbols expressed by an authority, in this case the IT sector acting as a collective intelligence.
There is not room for a real time new induction of meaning that is informed by individual and collective "measurement in the moment".
The OASIS BCM standard (recently approved) has a stratified ontology where invariants of past "measurements" are produced, aggregated or orchestrated, as "parts" of a blueprint, which is then offered to humans, or a human collective, for choices. A series of choice points may be produced using the BCM standard. These choice points then become where the origin of design shifts.
I do believe that the BCM notions of choice points and blueprints can only be improperly implemented into a SOA ontology using OWL.
The reason for the improperness is the same as I have argued. The ladening of RDF with description logic asserts too strongly a precise correspondence between symbol systems and natural reality. Sometimes the utility is good and the error not so pronounced; but there is no formal means to determine when the correspondence in a specific instance is incomplete or inaccurate. John Sowa makes a different argument related to the relationship between RDF and common logic.
The various small ontologies presented by Chris are all useful as a framework, a process model for “services”. I suspect that the SOA community may eventually adopt a SOA ontology, but one expressed as a topic map, as well as an OWL ontology. As long as the SOA community insists that there is no utility to topic maps, then this ontology will be like various bio-informatics ontologies (such as the biopax ontology for gene and cell expression). The biopax ontology is OWL full, btw, but is not standard OWL full as there were some modifications.
Any OWL SOA ontology will be used as an engine in cases where there is good clarity over that engine, and its deterministic functioning. Yea, good work!
However, as a general rule the discovery of aspects of situations cannot occur with the OWL assertions over category or class-subclass, particularly with the class of description logics. Chris's question about OWL Full and OWL-S reminds us, perhaps the reader, that there is still something missing in the OWL description of itself.
The problem is, I have claimed, that induction of meanings is situational or formal. Formal induction builds things like mathematics or a specific version of description logic, and does this in an artificial world of abstractions. In the real world of social interactions the induction of meaning is far more complex and richer.
Again, the point made by Joshua is reinforced by Brad's comment about class and objects. Measurement is made between real systems and involves activities at various levels of physical organization. For example, an agent concept may refer to a human being who has motivation.
The effect that the human being is motivated to achieve may be indirectly linked to aspects of the social world that are not modeled anywhere in the formal definition of the OWL call definition of effect. This is proper and is accounted for in the topic map standard, but not in the OWL standards.
Paul S Prueitt
The Taos Institute Org
Taos New Mexico
[1] OWL stands for “Web Ontology Language” and is the U.S. government’s funded solution to human knowledge representation. The funding agency decisions pretends to have looked carefully at all alternatives, including topic maps. See “White Paper: Resilience Project”, authored by Paul S Prueitt (2006).
URL: http://www.ontologystream.com/beads/nationalDebate/ResilienceProjectWhitePaper.htm