IRIs as node labels (proposals and counterproposals)

(changed subject line)

On Fri, 22 Mar 2002, Jeremy Carroll wrote:

> > 14: IRI's
> > Wheras nodes in an RDF graph are labelled with URI's and the
> > standards for internationalization of URI's are not yet stable
> >
> > Propose the WG:
> >
> >   1) resolves that nodes in RDF graphs are labelled with standard
> > web identifiers
> >
> >   2) resolves that the current standard web identifier is a URI
> > as defined by RFC 2396
> >
> >   3) resolves that resolution 2 above may be updated by an errata
> > to the specifications as new standards evolve.
> >
>
> This is unclear. First my counterproposal, second a critique of the chair's
> proposal.
>
> Counterproposal.
>
>  Propose the WG:
>
>    1) resolves that nodes in RDF graphs are labelled with standard
>       web identifiers
>
>    2) resolves that the current standard web identifier is a URI
>       as defined by RFC 2396, in its original character sequence
>       (in UTF-8)
>
>    3) notes that the use of identifiers of resolution 2 that are
>       not in normal form C presents internationalisation
>       difficulties and security risks
>
>    4) resolves that resolution 2 above may be updated by an errata
>       to the specifications as new standards evolve.
>
>    5) that equality between the identifiers of resolution 2 is
>       binary identity.

I have a counter-counter proposal:


(this goes beyond the IRI issue, and is motivated in part by my
exploration of the SOAP 1.2 Encoding Data Model, which uses XSD datatypes
but doesn't label nodes with URIs)

Propose the WG:

1) resolves that Web identifiers, like other characteristics of a
   resource, can be considered properties of the the resource that they name

2) notes that RDF now provides datatyping facilities which can use
   XML Schema datatypes (such as xsd:anyURI)

3) notes that RDF currently provides a privileged role for URI referring
   expressions. Nodes in an RDF graph may be labeled using RFC2396 URI
   references to indicate the resource that they represent.

4) further notes that this same information can be represented in a number
   of ways using RDF properties, and that at this time RDF Core provide
   no specific guidance on the representation of URI node labels as RDF
   properties, or on equivalencies between these two strategies for
   representing web identifiers in RDF.

5)
   resolves that future revisions to RDF could accomodate IRI referring
   expressions using a named property or datatyping convention.

6)
   notes that labeling RDF graph nodes with new kinds of referring
   expression (for eg. RDF definite descriptions) might be proposed as a
   design feature for any hypothetical RDF 2.0 effort. The simpler case
   of URI resource identifiers can be addressed using RDF's existing
   property and datatyping machinery.




Hmm, if this is true I ought to my money^H^H^H^H^Htestcases where my
mouth is.

I also have another argument (related to reification) in favour of our
doing URIs as properties, but it'd be a distraction to include here
(and probably a distraction from shipping Core in a timely fashion).

Dan



-- 
mailto:danbri@w3.org
http://www.w3.org/People/DanBri/

Received on Friday, 22 March 2002 06:03:52 UTC