W3C home > Mailing lists > Public > w3c-rdfcore-wg@w3.org > November 2002

Technical tweaks to the MT, for reviewers.

From: pat hayes <phayes@ai.uwf.edu>
Date: Fri, 8 Nov 2002 11:47:33 -0600
Message-Id: <p05111b23b9f19fd70c82@[65.217.30.130]>
To: w3c-rdfcore-wg@w3.org
Cc: pfps@research.bell-labs.com, herman.ter.horst@philips.com


1. There is a problem I mentioned earlier with literals like 
"badnumber"^^xsd:integer . They have to denote something, but they 
shouldn't denote a literal value (because they, er, don't, right?) 
The trouble is that until we go ask the datatype, RDF can't tell the 
difference between one of these and "12345"^^xsd:integer. So in the 
basic RDF MT, we need to say that typed literals denote something but 
we don't yet know even if its a literal value, so it has to be in IR, 
not in LV (as at present).

So what do we do when we find out that it is bad? The current draft 
invents an ad-hoc thingie (a triplet) to be the value in this case, 
but that is not a good solution. Better would be to say that it 
denotes some unknown but arbitrary thing which, whatever it is, is 
not a literal value. This adds a slightly odd technical extension to 
the definition of datatyped interpretation, which will now include a 
new mapping IBL from datatyped literals to things in IR that are 
*not* in LV, which are the 'things' that the literal will denote *if* 
it is badly typed. So, just a warning, I am going to make this 
change. It actually makes the semantic rules easier to state and the 
Lbase translation more transparent.

2. Now that subClassOf and subPropertyOf have 'iff' semantics, that 
raises a slight complication: it means that anything with an empty 
class extension, which right now includes anything that isn't a 
class, is a subclass of anything, and similarly for properties. That 
has a lot of unintuitive consequences, and I propose to remove them 
by *defining* classes to be things that are in  rdfs:Class, similarly 
properties and rdf:Property, and then restricting the 'iff' 
definition of subClassOf to hold only between classes, and again 
similarly for properties.

Summary of fix:

IP is defined to be {x: <x, I(rdf:Property)> in IEXT(I(rdf:type)) }
IC defined to be {x: <x, I(rdfs:Class)> in IEXT(I(rdf:type)) }

aaa rdfs:subClassOf bbb is true iff ///I(aaa), I(bbb) are both in IC 
and ///  x in ICEXT(I(aaa)) implies x in ICEXT(I(bbb))
and similarly for subproperty

The ///-/// marks the fix. This should not change any of the current 
rdfs closure rules (though if it wasn't done they would need 
changing.) It should have been done this way from the start, so this 
is a technical fix of a bug, in effect, and I apologize for not 
incorporating it into this draft in time.

3.  Recent changes, including the above, require several changes to 
some of the lemma proofs; these will be fixed. Herman ter Horst has 
kindly checked through the proofs in detail and made several good 
suggestions.

4. Does everyone agree that   rdfs:member rdf:type 
rdfs:ContainerMembershipProperty .  ? I can go either way, but this 
will require a small tweak.

Pat



-- 
---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola               			(850)202 4440   fax
FL 32501            				(850)291 0667    cell
phayes@ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
s.pam@ai.uwf.edu   for spam
Received on Friday, 8 November 2002 12:47:14 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 3 September 2003 09:54:00 EDT