Re: Agenda/Logistics, Oct 2 telecon

From: Jeremy Carroll <jjc@hpl.hp.com>
Subject: Re: Agenda/Logistics, Oct 2 telecon
Date: Wed, 1 Oct 2003 15:59:14 +0300

> Hi Peter

[...]

> I have a question on
> [[
> C/ Semantic layering bug
> 
>    The change to allow optional ontology type triples breaks semantic
>    layering at bit more.  (It was sort of broken in this area, but I think
>    everything might have worked anyway, due to some fortuitous
>    circumstances.) 
> 
>    Fix:  1/ Extend OWL abstract consistency and entailment to also allow
>             axioms and facts outside of ontologies.
>          2/ Extend OWL satisfaction of an abstract ontology to require that
>             the ontology be of type owl:Ontology.
>          3/ Extend the mapping to triples to allow for axioms and facts
>             outside of an ontology. 
>          4/ Make the typing triple mandatory for unnamed ontologies.
> 
>    RESOLUTION:  Approve the above fix.  The only documents that should need
>                 changing are S&AS and Test.  I accept an action to make the
>                 changes to S&AS.
> 
>    NOTE:  This effectively replaces the change of 18 September that made
>           rdf:type triples optional for unnamed ontologies, but gets the
>           same effect by allowing axioms and facts outside of ontologies.
> ]]
> 
> I wondered whether changing section 5 to require owl:Ontology to be non-empty 
> might be simpler.

I don't know if this works.  I don't like it as a solution in any case.

> Also I find the above proposal a bit of a pig-in-a-poke, i.e. I don't pretend 
> to understand it yet, and suspect I would need to see more fully developed 
> text to decide whether I like it ot not .... 

The basic idea is simple, I think, although the proposal above is a bit
difficult to decipher because I tried to make it too operational.

The basic idea is to allow ``abstract ontologies'' to possibly not be
ontologies at all, but instead just axioms and facts.  This is really what
many OWL documents are anyway.  So

	Class(Foo)
	DatatypeProperty(age)
	Individual(john type(Foo) value(age "25"^^xsd:integer))

would be top-level legal OWL abstract syntax, as would 

	Foo rdf:type owl:Class .
	age rdf:type owl:DatatypeProperty .
	john rdf:type Foo .
	john age "25"^^xsd:integer .

fixes 1, 2, and 3 are the changes that are required for this.

Fix 4 is to require that 

    Ontology(
	Class(Foo)
	DatatypeProperty(age)
	Individual(john type(Foo) value(age "25"^^xsd:integer))
    )

has a different mapping to triples than 

    Ontology(
	Class(Foo)
	DatatypeProperty(age)
	Individual(john type(Foo) value(age "25"^^xsd:integer))
    )

and, I believe, gets semantic layering back.

> We agree on (E) at least - well I am happy with the suggested tests (I have 
> already added them) and that will do on that one.

Are these new tests available?

> Jeremy

peter

Received on Wednesday, 1 October 2003 10:13:39 UTC