Re: Possible small RDDL enhancement

[Jonathan Borden <jborden@mediaone.net>]

Simon St.Laurent wrote:


> At 10:42 PM 3/18/01 -0800, Tim Bray wrote:
> >It dawns on me that it might be useful for a RDDL to declare
> >what namespace or mime-type it claims to be describing.  Would
> >it make sense to create a new optional element type, <rddl:class>,
> >with attributes ns= and/or media-type=, which could be used for
> >this purpose?  I was thinking of using attributes, but sometimes
> >a rddl might be used to describe more than one class of documents.
>
> I got into that a bit in XPDL:
> http://www.simonstl.com/projects/xpdl/wd051000.html#Sec2.2

It is interesting how the same issues and solutions keep popping up. I must
admit that I totally missed this first (and second time around) -- of course
I was expecting stuff to do with "processing" rather than defining what a
namespace ought be, but of course as is often said: one person's processing
is another person's data.

XPDL does clearly lay the groundwork for 'document processing as a function
of namespace dereferencing'
http://www.simonstl.com/projects/xpdl/wd051000.html#Sec3.7 though as you
note things like media-types (and for that matter notations) can be useful
in some situations but are generally broken as a robust way to associate
documents and processing. I think we can do better and agree with you below
that this is good area to explore. The bottom line will be demonstrations
and working implementations. I hope that people use RDDLClassLoader (java)
but especially develop similar functionality for other languages.

>
> I'm also really intrigued by the possibility of interacting descriptions,
> where it becomes possible to talk about multiple namespaces in a given
> document in some coherent way.

Something to mention is that 'xml:base' is very similar to 'rdf:about' in
the sense that setting the xml:base of a rddl:resource sets -what- the
rddl:resource is referencing. For example this might be a way to indicate
that the document is intended to describe these namespaces:

<rddl:resource
        xml:base="http://example.org/namespace1"
        xlink:arcrole="...purposes#description"
        xlink:href=""
><p>We are talking about this namespace -1-</p>
</rddl:resource>

<rddl:resource
        xml:base="http://example.org/namespace2"
        xlink:arcrole="...purposes#description"
        xlink:href=""
><p>We are also talking about this namespace -2-</p>
</rddl:resource>

which translates into the RDF:

<rdf:Description rdf:about="http://example.org/namespace1">
    <purposes:description rdf:resource=""/>
</rdf:Description>

<rdf:Description rdf:about="http://example.org/namespace2">
    <purposes:description rdf:resource=""/>
</rdf:Description>


> I'm especially intrigued by the RELAX
> 'divide-and-validate' bit:
> http://www.xml.gr.jp/relax/divideAndValidate.html
> http://xmlhack.com/read.php?item=1109
>
> I'm starting to think about programs that explore RDDL files over the
> course of processing documents.  Could be some interesting possibilities
there.
>

Leigh Dodds suggested the same thing for 'aggregating' Schematron schemata
for RSS having multiply namespaced contents/modules.

This is along the theme of "schema at namespace URI" except that various
schemata are thus *properties* of the namespace (which is a good thing IMHO)
rather than a particular *equated* to the namespace (which would be a bad
thing IMHO). In this model a namespace is indeed like a class whose
properties are the resources the RDDL document contains.

-Jonathan