RE: RDF Datatyping Working Draft

>Looks like a promising start ...
>
>I feel that your pictures can help me clarify my objection:
>
>Section 6.1.2 RDF Model Theory Interpretation
>http://www-nrc.nokia.com/sw/RDFDatatyping.html#ntoc_26
>
>and
>
>Section 6.1.3 RDF Datatyping Interpretation
>http://www-nrc.nokia.com/sw/RDFDatatyping.html#ntoc_27
>
>are different.
>
>In 6.1.3 all the properties have the same values (satisfying "Jeremy's
>entailment"). In 6.1.2 they don't.

Well, they might. There is no way to tell, since the values are 
denoted by blank nodes. But that isn't enough to make Jeremy's 
entailment hold, indeed.

>
>Now, if we leave it at that, and have no model theory for datatypes, then
>that seems fair enough, and much of my objection is satisfied. But if the
>model theory addresses half of datatyping but another half lies somewhere

No, look. The MT addresses *how* datatype constraints are imposed on 
an interpretation. It does that quite precisely (section 5). What it 
does not do is to say *what the particular datatype constraints are*, 
since that of course depends of the datatype itself and hence on 
things outside RDFS. So in any given example, the information about 
the particular datatypes involved has to come from somewhere else; 
but there is an MT equivalent of an API for the information: it has 
to specify membership in a lexical domain and a lexical-to-value 
mapping. Once that is done, the MT handles the rest by imposing 
appropriate semantic conditions on interpretations. So in a sense, 
the MT handles all of datatyping, as far as RDF is concerned. It 
completely specifies the way in which (information about) a datatype 
can impose meaning on an RDF graph. I don't see how it could possibly 
do more than this.

>else it feels very unclear as to when one should use a model theoretic
>interpretation and when one should use a datatyping interpretation.

I think the terminology is misleading here. They are both 
model-theoretic interpretations. Adding datatyping constrains the set 
of satisfying interpretations, is all., by imposing some extra 
conditions on satisfiability.  6.1.2 is the case where there is no 
datatyping, 6.1.3 shows the effect of adding the datatyping 
constraints to the interpretation.

>It also feels dubious that the relationship between model theoretic
>interpretations and datatyping interpretations is non-monotonic.

It ISNT nonmonotonic!   There is extra information in 6.1.3, but its 
not inconsistent with 6.1.2. The blank nodes can denote anything, 
remember, so 6.1.2 is underdetermined at those nodes, while 6.1.3 
fixes their values by imposing the datatyping conditions on the 
interpretation.

>(And that
>this non-monotonicity seems to be motivated primarily by the desire for tidy
>literals - my valuation is that non-monotonicity is a greater evil than
>either tidiness or untidiness).

I don't think it has anything to do with tidyness.

>I would be satisfied by:
>- no MT for datatypes
>- non-commitment over tidy literals
>
>(I believe we could have done better but ...)
>
>
>DO NOT feel obliged to respond to my points now.

Just make sure you get it right, is all :-)

>As far as I am concerned I
>lost the critical vote and was in a minority of one, and was hence rightly
>overruled. As the relevant WDs appear I intend to draft a dissenting opinion
>and circulate it to the WG and to my HP colleagues. After that, assuming my
>issues remain unaddressed, this would go on record as either my personal
>dissent or as HP's dissent. We have talked about datatyping enough.
>
>Jeremy
>
>
>Patrick:
>>  I just put a copy of the RDF Datatyping WD at
>>  http://www-nrc.nokia.com/sw/RDFDatatyping.html
>>  for those who would like to see where I'm at
>>  with it.


-- 
---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Wednesday, 3 April 2002 14:10:12 UTC