RE: how does existing RDF software handle this datatypes test?

DanC:
>
> If you read RDF/N3, and you don't mind some ommitted
> prefix declarations, you might find this short version
> easier to grok:
>   _:f rdf:type <#Film> .
>   _:f dc:title "10" .
>   <#mary> ex:age "10" .
>
> Suppose I asked you:
> 	does that document say that
> 	there's something with a dc:title?
> I hope you'd say: yes, of course.
> Now... suppose I ask:
> 	whatever that title is, let's call it X.
> 	Does that document also say that there's
> 	something with ex:age of that same X?


Now, in the discussion on RDF Core you have proposed thinking about this
question as an entailment question.

http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Jan/0440.html

[[[ DanC's message:
It's simpler than that; here's the premise:

  _:f rdf:type <#Film> .
  _:f dc:title "10" .
  <#mary> ex:age "10" .

here's the (purported) conclusion:

 _:x dc:title _:y.
 _:z ex:age _:y.
]]]


OK.
The only standards track document that addresses this is the RDF Model
Theory.
http://www.w3.org/TR/2001/WD-rdf-mt-20010925/

Both the current version and the soon to be approved next version agree that
bNodes cannot bind to literal nodes (see
http://www.w3.org/TR/rdf-mt/#unlabel ). (Pat has used some very obscure
wording (at least to me), I fear the conclusion may hinge on whether
literals are resources or not !?! The latest editor's draft of the model
theory http://www.coginst.uwf.edu/users/phayes/w3-rdf-mt-current-draft.html,
with untidy literals is clearer for this entailment).

Hence the (purported) conclusion is not entailed under the RDF Model Theory.

You are correct to say, that under TDL the entailment given does not hold.
(Even though in TDL, unlike the current model theory working draft, bNodes
can bind to literal nodes).

That is, TDL is backwardly compatible with the current model theory working
draft.

===

Now, let's consider S.
Is S backwardly compatible with the current model theory working draft?

If under S bNodes can bind to literal nodes then the entailment holds and S
is not compatible.
If, on the other hand, under S, bNodes cannot bind to literals then the
single triple:

_:a rdf:value _:b .

is contradictory (by the semantics offered rdf:value by S-P), which is also
not backwardly compatible with the current RDF model theory.

(Note - S doesn't specify which of these is the case).

So I am concluding that TDL is backwards compatible and S isn't.

===

Dan -

 we only have working drafts and rough code in development, not fully
polished standards or fully deployed systems. There is no real backward
compatibility argument for query or entailment, but if you really want to
have one I think I can win it!

This argument really rests on a value judgment, "Is the title of the Bo
Derrick film the same as a child's age?", and the world seems divided on
that judgment.

Can't we just drop this one -- please, pretty, pretty please with a cherry
on top?

Jeremy

Received on Thursday, 31 January 2002 06:47:42 UTC