Re: RDF Semantics: partial review

>>  >>
>>>>"A name is from a vocabulary if ..."
>>  >>I hope that this can be simplified.
>>>
>
>....
>
>>I continue to think that the terminology 'from V'
>>as being different from 'in V' is very confusing.
>>It seems that people will typically identify the two
>>notions.
>
>.....
>
>>)
>>
>>I would suggest to omit the terminology 'from V'.
>
>I agree that the in/from/of contrast may be too delicate and is not 
>really needed in any case.
>
>Let me suggest that I adopt the following simplified convention, 
>which I think will be sufficient. A name (as now) is a URI or typed 
>literal. A vocabulary is a set of names. The vocabulary OF a graph is 
>the set of names that occur in the graph as the subject, object or 
>predicate of a triple. Interpretations are defined on a vocabulary, 
>usually that of a graph.
>
>Note, this excludes the URIs inside typed literals. Since IL applies 
>directly to typed literals, this will be of consequence only when we 
>consider datatyping explicitly, and in that case the requirement that 
>datatypes be 'declared' by a triple
>
>ddd rdf:type rdfs:Datatype .
>
>is sufficient to ensure that all the required URIs are part of the 
>graph vocabulary. (rdf:XMLLiteral is part of the rdfV vocabulary). So 
>I think in fact there is no need to even consider the names inside 
>typed literals when describing simple, RDF and RDFS entailment.
>
>This approach has the merits of simplicity and of treating all nodes 
>uniformly, which is more conventional in any case.
>

I agree 100%.


[...]

>>
>>>
>>>>==
>>>>
>>>>Section 1.5
>>>>
>>>>the table: Semantic conditions for blank nodes
>>>>
>>>>It seems that line 1 would need to be replaced by
>>>>something like the following more complete statement:
>>>>-  If A is a mapping defined on the blank node E, then
>>>>     I+A(E)=A(E).
>>>
>>>I don't feel that that is necessary.
>>>
>>
>>You cannot speak of A(E) unless it is defined.
>
>Oh, come. A is a mapping: A(E) is the result of applying that mapping 
>to E.  Do I really need to *define* what is meant by applying a 
>mapping to an argument? You will want me to be defining the meaning 
>of the word "and" next.
>

No: you do not need to define what is meant by applying a mapping 
to an argument.  And I will not want you to be defining the meaning of
the word "and" next.
The point is that A(E) can be undefined, since the text
specifies A to be only a mapping from "*some* set of blank nodes to
the universe IR of I".
On further reflection, I can give a suggestion for a much shorter
correction than the one I gave above:
   If E is a blank node and A(E) is defined then I+A(E)=A(E)

[...]

Herman

Received on Wednesday, 12 November 2003 09:06:26 UTC