Some talking points for today

IRC logs of a predicussion IRC chat twixt Tim, me, and some other folks  
on #rdfig.
	http://ilrt.org/discovery/chatlogs/rdfig/2003-10-08.html#T20-38-33

**THINGS PERHAPS APPROACHING CONSENSUS

1) Strong "Ontological Commitment" is not required. To wit, there's no  
MUST, or even SHOULD that your document is consistent if and only if  
the imports closure over URI *used* in your document (+ your document)  
is consistent.

There's a meaning-rathole version of this, or meaning-rathole  
consequences, but I'm avoiding them at the moment.

2) It may be interesting for programs to do (or respect the effect of)  
various sorts of imports closure.

	I'd be willing to upgrade that away from the "may" except that I think  
this is subsumed by, "it may be interesting for programs to do  
information gathering and to make choices about how they interpret  
third party assertions and how they incorporate them into their kb and  
how they act on them." I regard this as a truism, but that might be  
controversial.

Hmm. That's it. I guess. Any other candidates? Maybe:

3) *owl:imports* has some infelicities.

Perhaps better was around:
	http://ilrt.org/discovery/chatlogs/rdfig/2003-10-08.html#T20-56-54

4) Consistency of a RDF (or OWL) document consists in it being possible  
that all the statements asserted in that document can be co-true (to  
phrase it colloquially; for a crisp definition, see the specs or a  
logic text).

	This may seem trivial, but it's actually a fairly specific notion of  
consistency (glorious though it may be), and it wasn't always clear  
that everyone meant this.

	In the IRC Chat, Tim and I managed to *not* have me dyed with the  
specter of having defined a special kind of consistency for OWL and  
getting us down the road of a whole family of "consistencies". There  
may be an infelicity in owl:imports being in *OWL* rather than in *RDF*  
(i.e., that an RDF Parser shouldn't respect owl:imports). It *may* be  
worth making some sort of comment about that.

**THINGS NO WHERE NEAR CONSENSUS

5) The "priority of the predicate" to meaning.

6) The "meaningless" (or, at least, inutility) of RDF documents given  
the current specs.
	Do the RDF spec fail in their "duty", as Tim puts it?

7) That the current specs either suggest or partially entail 1, but not  
enough to avoid 2.

8) What kind, if any, infelicity a document having an inconsistent  
imports closure has. (And whether there needs to be spec language about  
this.)

9) The denotation, referent, meaning, or singularity of these of URIs.

10) What URI ownership *really* gets you.

11) The worthwhileness of pursuing specification of "naive protocols"  
for...something. Import closure or "coherence".
	Note: I'm a bit bummed by the "protocol" talk. I don't consider it to  
be an advance. There seems to be a bunch of baggage associated with the  
use in this discussion. In fact, a sorta rhetorical trick to try to  
level the playing field for the "protocol engineers". It was strongly  
introduced by Tim's reply to my issue framing (see my comment on this:  
http://lists.w3.org/Archives/Public/public-sw-meaning/2003Oct/ 
0005.html). So, it may be my personal baggage, but it still feels akin  
to the "WE MUST SPECIFY! YOU DON'T GET SPECS AND THE IMPORTANCE OF  
THEM!!!!" line. Let me state here that there is no need for and some  
disutility in this line. Certainly neither Peter, nor Pat, nor myself  
fail to appreciate specifications and what they can do. We disagree  
about so proposed *content* (and perhaps, some derivations of what  
should be or is in some specs' content).

	Note2: This item is not *just* about people who think we don't need  
any of these "protocols" in a spec vs. those who do, but about those  
who think we should jump straight to "robust protocols".

12) Weak "ontological commitment". You shouldn't use a URI in a way  
that conflicts with its owner approved definition. I'm not sure exactly  
how to distinguish this from 1, or 11, or 10. But the intuition for  
distinguishing it from 1 (and making it a kind of 11) is that with 1,  
you're dealing with the *WHOLE* kb at the other end of the imports. So  
any inconsistency anywhere infects your document. Here, there's  
something like the idea that you shouldn't conflict *only* in how you  
use that URI. Or something.

**THING TIM, DANC, AND I MAY HAVE CONSENSUS ON BUT IT'S UNCLEAR FROM  
THE IRC LOGS

http://ilrt.org/discovery/chatlogs/rdfig/2003-10-08.html#T21-48-11

Roughly, the modified null-hypothesis: We ask the TAG to close the  
issue and continue with the sw-meaning list (or a sucessor list)

**A META POINT

If we do have some consensus on one, it will be worth reporting back in  
some semi-official way to various groups. E.g., as a lc comment (a  
happy one!) to rdf core.

Received on Friday, 10 October 2003 09:24:22 UTC