Re: Tech Plenary: agenda Best Practices

>> Just that 'meaning stability'; what do you mean by that ??
>
> Well, its rather vague, but imagine that someone publishes an
> ontology of, say, aircraft, or people, or agent actions and policies,
> and it turns out to be useful so that the concepts in it get used by
> others. Now the originators realize that they want to change
> something. Strict W3C-grade Web etiquette requires that they make no
> change to the document at the published URL. But if they produce a
> new version at a new URL, all the old URIrefs in all the other
> ontologies will need to be upgraded. It seems more reasonable to
> think of a Web ontology as something more like a 'moving resource'
> which can get upgraded from time to time, have bugs fixed, be
> generally made better and so on, and still be the 'same resource' in
> the REST web-architecture sense, just with its state brought up to
> date from time to time. OK, but then what KIND of changes to an
> ontology are reasonable for it to be in some sense the 'same
> ontology' ? Im not sure, but I think the question needs to be
> addressed: and I don't think that 'no changes allowed' is a
> reasonable default. It ought to be more like "barring error
> correction, you can still infer the same stuff you could infer
> before" , maybe. Or perhaps there should be a category of 'moving
> ontology' (? Brentano would have liked this...) which when accessed
> gives you an ontological snapshot of the current state of something
> at the access time (the weather forecast rendered into RDF): those
> would obey different rules as they would be expected to change. (Or
> maybe not: maybe we ought to encourage such moving ontologies to
> invent new properties which once invented are eternal, such as
> ...#weather030420041646USCT rather than updating the value of
> #weather, so that a moving ontology is always seen as a growing
> ontology, at least conceptually(?) .)  Maybe there are other kinds
> that one would expect to obey still different rules.  And so on. In
> general, there ought to be something identifiable as 'the intended
> meaning' of the ontology which is stable in some sense and which is
> what one can reasonably expect a given URI to go on identifying.
> This certainly isn't the syntactic form of the document, and it isn't
> going to be the strict model-theoretic interpretation in all cases,
> either. I'm not sure what it is, but I have a hunch we need to figure
> it out well enough to write best practice guides to it.

I'm actually inclined to a bottom-up approach...
(grass doesn't grow by pulling at it...)
Taking common/proven cases and semwebize them shows
what kind of additional cases can be made explicit
and so "you can still infer the same stuff you could
infer before" but you could also have the benefit of
additional inferences.

--
Jos De Roo, AGFA http://www.agfa.com/w3c/jdroo/

Received on Saturday, 6 March 2004 21:18:05 UTC