Re: [WNET] new proposal WN URIs and related issues

Mark,

On Wednesday 19 April 2006 13:50, Mark van Assem wrote:
> I'm not sure what you think the motivation was... Actually thinking back
> I think the motivation was twofold

I guessed (2)

> 1) esthetics
> 2) easy way to define URIs that refer to larger chunks, e.g.
>   http://wordnet.princeton.edu/wn20/wordsense/bank could refer to all
> the wordsenses containing 'bank'. I moved this idea to the 'Issues'
> section of [1].
>
> > problem must be solved otherwise. Of course, unless you propose to revise
> > XML/RDF and preferably also XML namespace handling :-)
>
> I think especially that only allowing QNames for attributes in RDF/XML
> is pretty strange. You can't even replace <rdf:type ...> for its
> expanded URI. Probably other notations do not have this quirk?

You seem to forget the purpose of SGML/XML was designed for: define the
structure for a text document. Elements and attributes typically have
names like 'author', 'h1', 'name', etc. You `invent' these names while
defining the meta-structure of documents (DTD).  They are identifiers, just
like function and variable names in programming languages.  Just to
avoid complex quoting rules and unreadable identifiers they are limited
to -simplified- alphanumerical characters and a few things one can use
as word-breaks (-_.).  The '/' already plays a role in XML syntax, so
you don't want that to be in identifiers.

The RDF/XML serialization tries to represent an instance of a class
in a natural way as <Class name=value><name2>value2</name2></Class>.
This implies `Class' and `name' must represent URIs, but XML poses
some limits on this.  Alternatively one could have choosen for
something like this:

<rdf:triple subject="..." predicate="...">object</rdf:triple>

There is something to be said for both approaches. Overall, I'm not
unhappy with the current XML/RDF standard.  Just give all students
an excercise to write a parser for it :-)

	Cheers --- Jan

Received on Wednesday, 19 April 2006 12:40:16 UTC