Re: asn07

Dave Reynolds <der@hplb.hpl.hp.com> writes:
> Sandro Hawke wrote:
> > I've put up some details of a proposed revision to asn06 (called,
> > surprise surprise, asn07).
> >       
> >    http://www.w3.org/2005/rules/wg/wiki/asn07
> > 
> > The big differences are:
> > 
> >    - support for lists/sequences has been removed, as per my discussion
> >      with Bijan and others about how to model sequential items
> 
> Could you summarize those?

Bijan's basic argument (as I recall) was that things rarely have an
intrinsic order, but, rather, one puts them into a particular order for
a particular purpose.  cf databases, where the ordering is done at query
time.

> ASN07 is aimed at modelling data structures and an ordering construct is 
> pretty useful there. In particular how will the ARG list in Uniterm be 
> modelled without an ordered sequence construct?

Uniterm is one of the best candidates for an exception to the above
generalization.   :-)

Something like this should work, though:

    class Uniterm
       property op: Const
       property arg: Argument*

    class Argument
       property value: TERM

       subclass PositionalArgument
          property position: int

       subclass KeywordArgument
          property keyword: Const

It's more general -- it's probably the kind of thing you need if
supporting keyword arguments -- but I know it could also lead to an
absurd bloat of the XML syntax.   (I can picture the Monty Python skit
in my head right now.)

Hrm.

I don't like complicating and confusing RIF with this issue.  I just
want to be able to nail down the semantics of the asn.  I really like
the idea of having it be a sublanguage of OWL, but if it can't be, it
can't be.   I guess the "list of Term" approach is still kind of doable
in OWL (using either rdf:List in OWL-Full, or a new list vocabulary, as
you and I have discussed before).

> You use but don't define "qname" in the EBNF. Is that really a qname 
> syntax, as in the name part has to be a legal NCNAME?

That was my intent, I think.  Am I missing something?  Do you see a
reason for increasing the allowed characters?

> It would worth changing the examples to use qname syntax to be 
> consistent with the EBNF.
> 
> Might also be worth spelling out that you are using RDF-style 
> concatenation to convert a qname to a URI to make the OWL mapping work. 
> [At least I assume that's what is meant.]

Right.

     -- Sandro

Received on Tuesday, 29 May 2007 18:29:56 UTC