Toward concrete types [was: Joint...]

From: Dan Connolly (connolly@w3.org)
Date: 12/19/00


Mike Dean wrote:

> concrete types proposals from Mike [5], Dan Connolly, and
> Peter/Ian
[...]
> [5] http://www.daml.org/committee/minutes/2000-12-19-mdean-concrete-types.html

Thanks for getting the ball rolling, Mike.

I don't completely understand your proposal:

|Specifics
|Allow use of an XML Schema Datatype URI as the range of a property.
Such a |URI may be either 
|a reference to a builtin data type defined by the standard, e.g.
|http://www.w3.org/2000/10/XMLSchema#string 
|a reference to a user-defined derived type contained in an appropriate
XML |document type,

Which XML document types are appropriate? I don't see exactly
what you mean here.

| e.g. http://www.foo.com/2001/01/application#ssn 
|Such URIs will implicitly be considered subclasses of rdfs:Literal. 

What exactly does "Such URIs" mean? Is the following such
a URI?
	http://example.org/foo#bar

| DAML components

(I assume I can use "DAML component" and "DAML agent"
interchangeably.)

| may make use of the range constraint (and any other information
| derivable from it, e.g. by retrieving the referenced XML Schema
| definition), or may behave as if rdfs:Literal were specified instead
| (i.e. just treat the value as a string). 

Hm... I don't recommend requiring DAML Agents to grok XML Schemas.
It's straightforward to reflect the info from an XML Schema in RDF,
and it's straightfoward to publish both an XML Schema and
a DAML/RDF schema for a namespace, and use HTTP content
negotiation to pick which one you want. Or just use different
URIs, and have the content of one talk about the other.

So I'd rather use something like:

<rdfs:Class
about="http://www.daml.org/committee/minutes/2000-12-19-mdean-example.xsd#ssn">
  <rdfs:subClassOf
resource="http://www.w3.org/2000/10/XMLSchema#string"/>
  <xsd:pattern>\d(3)-\d(2)-\d(4)</xsd:pattern>
</rdfs:Class>

Many of the XML Schema facets can be modelled straightforwardly
as restrictions, I think... especially if we add restrictions
for hasValueLessThan etc. in addition to hasValue.

Also... your proposal mentions the use of concrete types
as the value of an rdfs:range property; but they're not
limited to that, are they? They can be the value of any property...
or the subject of a property, for that matter, no?

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
mailto:connolly.pager@w3.org?subject=pls%20call%20+1-NNN-NNN-NNNN


This archive was generated by hypermail 2.1.4 : 04/02/02 EST