On 9 March, Dan wrote:

> On Thu, 2006-03-09 at 21:59 +0000, Misha Wolf wrote:
  [referring to the draft TAG finding [2]]
> > Either that draft TAG finding is broken or I'm missing 
> > something.
> > 
> > > Then perhaps we do need another example.
> > 
> > I don't think that another example will fix it.  There seems 
> > to be a conceptual gap here.  Why does the finding have a 
> > section called "3.3 Using GRDDL", which starts  with: 
> > "A third approach is to use [GRDDL]"?
> >    ^^^^^
> > Why is this a separate "approach"?
> That strikes me as odd too, now that I look at it.
> I guess the point is: you're encouraged to put information so
> that people can look up namespaces in the Web. It's good to
> have both machine readable and human-readable information there.
> RDDL 1.x is one known way to do it. Some folks like RDDL 2.x
> better. GRDDL also does the trick.
> The finding doesn't (yet) explicitly cover the fact that RDDL 
> documents can also be GRDDL documents.

This is where the gears in my head produce a grinding noise.  The 
acronym "GRDDL" seems to be being used to describe a number of 
different things.  Though I've gone through the GRDDL draft [1] a 
number of times, I continue to be puzzled by some of the uses of 
the acronym, both in the draft TAG finding and in your mails.  I 
have no idea of what you mean by a "GRDDL document".  The GRDDL 
draft doesn't seem to contain such a term.

What I'm after is a widely-adopted mechanism which will enable
machines to extract the underlying RDF from NewsML 2 documents.
The NewsML 2 WP struggled with how to reconcile our desires to 
(i) place NewsML 2 squarely within the Semantic Web, and (ii) 
avoid the RDF/XML syntax.  When, in May 2005, we heard of GRDDL 
(in Steven Pemberton's Keynote to the News Standards Summit 2005) 
we were delighted.  Since then, though, the rather murky GRDDL 
picture seems to have stayed rather murky.

We need to be able to associate multiple resources, such as XML 
Schemas, documentation, etc, with (the class of) NewsML 2 documents.
RDDL seems to be a good match for this need.  We then also want to 
be able to specify a transform which will convert a NewsML 2 
document to some flavour of RDF.  The most obvious solution would 
be to locate the link to this transform in the same RDDL document.

There is a twist to this, which is that the NewsML 2 Schemas will 
have extension points, allowing providers to extend the language 
to cater for requirements which have not (or not yet) been accepted 
by the IPTC.  I imagine that this is where we would need to utilise 
the "transformation" attribute, specified in section 3 of the GRDDL 
draft, to enable the provider to link directly to a transform able 
to understand their extended NewsML 2.

> >   Why doesn't the finding 
> > simply offer a Purpose such as "Serves the purpose of conversion 
> > to RDF/XML"?
> >
> > > To give a link from a namespace to a transformation
> > > for documents that use that namespace, GRDDL coins
> > > this URI:
> > >
> > 
> > So why isn't this included in the list of Purposes in the draft 
> > finding?
> Good question. That list of purposes is sorta odd.
> We're having trouble finding the end of it.
> In some sense, every RDF property URI is a perfectly
> good Purpose; clearly we can't list them all, but
> if we're going to list the RDDL ones, we might
> as well list the GRDDL ones too.

Ouch!  The gears are making that grinding noise again.  What do you 
mean by "the RDDL ones" and "the GRDDL ones".  How can there be RDDL 
ones?  And how can there be GRDDL ones?  RDDL is simply a mechanism 
for encapsulating typed links.  And GRDDL provides some typed links.
So I suppose it is meaningful to speak of "the GRDDL ones" as being 
typed links to resources which do transforms from ANOtherML to 
someFlavourOfRdf.  But when the draft TAG finding and your mails 
present RDDL and GRDDL as being on the same plane, my head hurts.

> > > GRDDL also coins 2 other URIs that can be used as RDDL
> > > purposes:
> > >
> > >
> > 
> > Ditto.
> > 
> > > > [1]
> > > > [2]

------------------- NewsML 2 resources ------------------------------        | |

To find out more about Reuters visit

Any views expressed in this message are those of the individual sender, except where the sender specifically states them to be the views of Reuters Ltd.

Received on Friday, 10 March 2006 13:40:53 UTC