- From: W. E. Perry <wperry@fiduciary.com>
- Date: Sun, 04 May 2003 12:20:24 -0400
- To: "Roger L. Costello" <costello@mitre.org>
- CC: www-rdf-interest@w3.org
Hi Roger. Undoubtedly you know what I am going to say to this, but FWIW here it is: The point of a semantic web is to enable networked processes to share or to interchange data and, more particularly, conclusions from data and logical relationships expressed as taxonomies. The overarching criteria in implementing this are to afford each process the advantages of leveraging for its own purposes the taxonomies assembled at every other node of the web and to achieve universal scale in that undertaking. By design, this formulation does not accommodate agreement on definitions in more than the most general terms. Your colleague correctly notes the entropy as definitions evolve independently over time at autonomous nodes of this web. Yet such autonomy with respect to definitions is necessarily the general case, even without consideration for separate local changes over time. The motivation of this web is to serve the individual local processes which comprise it. Each such local process requires data--and the structure, both logical and physical of that data--to be utterly specific to the particular needs and expectations of the process if it is to be of any use to that process. In other words, local requirements of process, including locally-specific definitions and ontological understandings, have priority over shared semantics or understandings-in-common because the point of the web is to harness for the benefit of the whole the expertise of specialized local processes. To the extent that those processes are indeed specialized, they must depend upon and indeed prefer locally-specialized definitions of data and locally-understood ontologies over those aspects of data definition and of ontological taxonomy which they might share with other processes. Paradoxically, perhaps, this is not a weakness of the semantic web but actually the source of its robustness, scalability, and value in harnessing otherwise diffuse expertise. The apparent option to accept a priori single, inflexible definitions and canonical ontologies is simply not a design choice if we are building a web of specialized processes rather than cloning the implementation of canonical processes for elaborating agreed semantics. The needs of the local processes must come first because making the specialized expertise of those processes widely available is the point of the web. A priori agreement and expected semantics necessarily yield to idiosyncratic definitions and peculiar taxonomies because the autonomous processes which the web is designed to support require peculiarity of perception rather than definitions in common. Respectfully, Walter Perry "Roger L. Costello" wrote: > Hi Folks, > > A colleague sent me the below message. How would you respond to it? > > > ... a weakness in the Semantic Web, which is to say that there > > needs to be universal agreement on definitions or the process > > breaks down. Even if there is universal agreement at a point > > in time, definitions will evolve and mutate, as in regular > > language. > > /Roger
Received on Sunday, 4 May 2003 12:20:30 UTC