RE: semantics and RDF(S)

On Fri, 12 Jan 2001, Dickinson, Ian J wrote:

> Dan,
>
> From: Dan Brickley [mailto:danbri@w3.org]
> > The thing to clean up first is the general notion of URI
> > naming on the Web
> Maybe we should get everyone to drive on the same side of the road while
> we're at it? :-)

That's a very real concern. Precision in RDF w.r.t. resource naming and
identity requires precision in the  specs that RDF needs to interoperate
with (URI, HTTP, XML stuff etc). Even doing an analysis of all the specs
would be a tough job, let alone trying to get agreement on a shared model.

>	Seriously, while I can see the point you're making, isn't
> it the case that the web is always going to be messy?  It seems to me that

That's the story: it was messy, but it worked, whereas prior purist
approaches stayed in the lab. The current question is about how feasible,
desirable... etc it is to attempt a purist formal cleanup of Web
architectural principles. And the specs that use them.

It may well be that the Web will always be messy. In which case we'll
shrug shoulders and use RDF for sitemaps, RSS newsfeeds, Dublin Core,
thesaurus services, MP3 ripping and discovery, all that low tech stuff we
originally promised it for. Some of us hope we can go better than this,
but the shpae of the problem isn't really understood: are we trying to fix
RDF, or fix the Web? [discuss ;-]

> the web is unlikely ever to reach a state of rectitude according to clean
> theoretical principles, so tying success to that state could hold a lot of
> good work up.

which is why I spent much of my free time last summer working on low-tech
things like RSS rather than rdf-logic, even though the latter is just as
interesting. We have unfinished business delivering on some simple apps
(sitemaps etc); I'd be dismayed if RDF-Logic, DAML/OIL/SW/RuleML etc
efforts delayed still further these more modest goals.

> A further point is that ontologies have many more uses than for reasoning
> about web resources.

I meant reasoning in a broad and rather colloquial sense, that would
include for eg. the workings of a query planner trying to answer an
RDF-based question as speedily as possible, through use of DAML-like
Schema annotations.

			 Indeed wasn't DAML born from a desire in the DARPA
> agent research community for a common language for agents to advertise and
> discover capabilities?

...and reason about them :)


			So could there, in principle, be a version of RDF
> (or something like it) that had a clear meaning as the basis for ontology
> designers, and a messy or less-than-perfect match to the "meaning" of the
> web?

Well, that's sort of what we've had, and we've suffered from the confusion
_anyone_ faces when trying to implement around the notion of resource.
It boils down to very practical questions: people ask what tables they
need to create in MySQL to store URI names for resources, whether every
resource 'must' have a URI, how to represent trivially different uris etc
etc. But I agree (if I get your point) that we need a way to proceed
cautiously without waiting for some grand unification theory of Web
architecture.

Dan

Received on Friday, 12 January 2001 10:58:02 UTC