Re: DAML: restricting number of elements in a list

On February 11, Ken Baclawski writes:
> On Mon, 11 Feb 2002, Ian Horrocks wrote:
> 
> > On February 11, Ken Baclawski writes:
> > > On Sat, 9 Feb 2002, Ian Horrocks wrote:
> > > 
> > > > On February 7, Ken Baclawski writes:
> > > > > The daml:item property relates a list to each of its elements.  One can
> > > > > impose a daml:minCardinality restriction on the daml:item property to
> > > > > ensure that the number of vertices in a polygon is at least 3. 
> > > > > 
> > > > > Ken Baclawski
> > > > > Ken@Baclawski.com
> > > > > UBOT Project
> > > > 
> > > > I am afraid that this is a common misconception. You can write such a
> > > > thing, but if you read the language specification you will find that
> > > > daml+oil does not provide any semantics for it, so it does not have
> > > > the effect that you desire (in fact, from the point of view of
> > > > daml+oil, it has no effect whatsoever). 
> > > 
> > > You are correct that the intention of daml:collection is to be unordered,
> > > but the axiomatic semantics specifies that DAML lists do have an order,
> > > via the daml:first and daml:rest properties.  Furthermore, the axiomatic
> > > semantics specifies that daml:item relates a list to each item in the list
> > > (item-axiom-2).  Which one specifies the semantics of DAML+OIL:  the
> > > axiomatic semantics (written in KIF) or the language specification
> > > (written in English prose)? 
> > 
> > It isn't a question of whether lists are ordered or not (although as
> > you point out lists are treated as sets DAML+OIL), it is the fact the
> > the list construction forms part of the syntax specification of the
> > language. If you check he model theoretic semantics (the definitive
> > specification of the meaning of DAML+OIL), you will see how this
> > works. 
> 

The model-theoretic semantics was the original formal specification of
the language. The KIF axiomatization was based on the model-theoretic
semantics and should exactly reflect it (we hope). As you point out,
however, the model-theoretic semantics also includes an axiomatisation
for lists and for RDFS in general. You are right to say that the
precedence of definitions should be made clearer.

Ian

> Where does it say that the model-theoretic semantics is the definitive
> specification of the meaning of DAML+OIL?  In the Reference Description it
> states: "Two references that give a precise definition of the meaning of
> the language constructs are the model-theoretic semantics and the KIF
> axiomatization."  I could not find a statement in the Reference
> Description that the model-theoretic semantics takes precedence over the
> KIF axiomatization, nor a statement that either one is the "definitive
> specification".
> 
> It seems that there are at least 4 specifications of DAML+OIL, and none of
> them is fully consistent with any other:
> 
> 1. Reference Description
> 2. Revised Language Specification
> 3. Model-Theoretic Semantics
> 4. Axiomatic Semantics
> 
> For example, the Axiomatic Semantics formally specifies lists, while the
> Model-Theoretic Semantics handles lists informally.  In the Axiomatic
> Semantics and the Revised Language Specification, the daml:List class
> participates in the class hierarchy (e.g., List-axiom-1), while in the
> Model-Theoretic Semantics daml:List, daml:first, daml:rest and daml:item
> are not even mentioned.  The Reference Description states that lists are
> unordered, but this is not reflected in any of the other specifications.
> 
> It isn't clear to me how this situation contributes to a "clean and well
> defined semantics" as claimed by the Reference Description.
> 
> > The confusion is caused by the fact that RDF does not separate
> > modelling layers, so the "meta-model" (i.e., the specification of the
> > syntax of the language) is mixed up with the model (an ontology
> > defined using the language).
> 
> IMHO, it doesn't help matters for one of the specifications to be
> independently "correcting" this problem while the other specifications do
> not.  Whatever solution is chosen for this problem should be consistently
> reflected in all of the specifications.
> 
> Ken Baclawski
> Ken@Baclawski.com
> UBOT Project
> 
>

Received on Thursday, 28 February 2002 07:43:01 UTC