RE: Summary of the QName to URI Mapping Problem

> >
> > ...
> > >
> > >Now, we have here two functional layers: RDF, and HTTP. A SW agent
> > >may interact with that URI at either level, and the semantics at
> > >one level does not have significance at the other.
> >
> > Ah, I see what you mean, and agree with you. I thought (too hastily -
> > is there a smilie for eating humble pie?) you were saying something
> > larger and more, er, philosophical about the nature of meaning.
>
>I'm sure I'm greatly to blame for using imprecise terms, not being
>sufficiently well versed in KR or logic terminology.
>
> > DAML+OIL, the next 'layer' above RDF for some folk, actually does
> > make some nontrivial use of the HTTP semantics, in a sense, in that
> > it has ways of using URIs to refer to other sets of axioms and import
> > them. This is admittedly a rather simple kind of cross-layering
> > interaction, but it hints at the kind of issues that are going to
> > arise.
>
>But is this really an interaction of the semantics of the RDF graph
>and the semantics of the URI? I.e., aren't such mechanisms really just
>instructions to the parser?

You can look at it that way, indeed, and that might be the easiest 
way for just this. I guess my point is really only a kind of gut 
feeling/worry that we are going to have to be sensitive to ways in 
which the classical syntax/semantic layerings are going to get more 
complicated. Here's another example, from the RDF M&S (which has all 
kinds of wierd stuff in it, by the way, that seems to be very little 
used): section 3.4, aboutEachPrefix, which takes a URI prefix and 
asserts something about all the things that are on the web that can 
be retrieved by adding some extension to it. We didnt even try to 
give this a formal meaning in the RDF model theory, but in order to 
do so one would need to 'cross layers' rather drastically (not to 
describe the result of 'expanding' aboutEachPrefix, thought of as a 
preprocessing step - the current workaround - but to treat it as a 
genuine language element and give it a coherent formal semantics.)

I suspect that these 'workarounds' are going to start getting too 
much trouble after a while.

>Do those "include" instructions actually
>have explicit representation in the graph?

Well, yes, in the RDF graph of a piece of DAML. But to be fair, DAML 
itself doesnt claim to be a graph-based language.

>And if so, then does not
>DAML then define not only an extension of RDF semantics but also
>define a processing model providing methods of knowledge syndication?

Good question. I think the 'official' answer would be that the 
semantics - model theory - of DAML was the primary source of meaning, 
and indeed it does extend the RDF semantics in a sense - DAML is more 
expressive than RDF - and that DAML processors are free to do what 
they like as long as it is valid with respect to the semantics. So 
there isnt really a single definitive processing model, more a kind 
of semantic constraint that any such processor should satisfy.

>And is then the semantics of that processing model part of the semantics
>of the graph model? (sorry for so many questions at once)

No, I'd say not, even if there was one. But my word isnt law, so 
others might disagree.

>I still don't see how these layers interact in a way that the semantics
>at one level is "semantics" of the same sort or resolution at another
>level.

Well, natural languages are like this. Meanings that are expressed in 
one language in explicit words are conveyed in another language by 
the use of lexical variations (eg word endings rather than 
prepositions), so that if one were to seek a sharp distinction 
between, say, the lexical level - to do with word forms - and the 
syntactic level - to do with word arrangements - then one would find 
that a full semantic theory would be level-crossing in the same 
sense. Not, I hasten to add, that I want to be seen endorsing the use 
of natural langauge in these contexts, only to illustrate the general 
point.

> > >I.e. No URI Scheme can introduce any semantics into RDF. The
> > use of any
> > >URI Scheme for resource URIs has no relevance whatsoever to semantics
> > >associated with an RDF graph. Right?
> >
> > Right, other than ensuring the coherence of the assumption of global
> > name uniqueness (which I think is a *syntactic* matter for RDF.)
>
>Agreed. And in fact this whole mapping problem has been seen from the
>start as a syntactic matter (so why is it on the rdf-logic list? good
>question ;-)

Oh my God, is that where we are?? I'd better go change my shirt.

Pat Hayes

---------------------------------------------------------------------
(650)859 6569 w
(650)494 3973 h (until September)
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Thursday, 30 August 2001 19:37:47 UTC