W3C home > Mailing lists > Public > w3c-rdfcore-wg@w3.org > November 2001

RE: Issue rdfms-abouteach

From: <Patrick.Stickler@nokia.com>
Date: Fri, 16 Nov 2001 13:47:03 +0200
Message-ID: <2BF0AD29BC31FE46B7887732114404316217DB@trebe003.NOE.Nokia.com>
To: danbri@w3.org
Cc: jjc@hplb.hpl.hp.com, dave.beckett@bristol.ac.uk, w3c-rdfcore-wg@w3.org
> > <rdf:Description rdf:aboutEach="#guys">
> >    <rdf:type rdf:resource="x:Male"/>
> 
> We can't put a qname 'x:' there, so this would have to be
> 
>     <rdf:type rdf:resource="http://example.org/schema#Male"/>

Yes, yes. Sorry. Typing in a hurry. I'd normally put

     <rdf:type rdf:resource="&x;Male"/>

and define the URI prefix as an XML ENTITY...

> ...slightly less prettier than doing it inline with a typedNode:
> 
>     <rdf:li><x:Male rdf:about="#phil"/></rdf:li>
> 
> ...that said, your general point is fair: some things (eg. ascribing
> further properties about each item) would be prettier/easier. 

Well, rdf:type was just *one* property that might be ascribed, and
the fact that there is this alternate "inline" type of definition
perhaps makes it a poor choice for example. Also, I may wish to
define numerous common properties to those bag members, e.g.

  <rdf:Description rdf:aboutEach="#someID">
     <x:property1>value1</x:property1>
     <x:property2>value2</x:property2>
     <x:property3>value3</x:property3>
     <x:property4>value4</x:property4>
     <x:property5>value5</x:property5>
  </rdf:Description>

And you can't do that otherwise in any convenient manner, other
than enumerating all those statements for every single resource.

Eh?
 
> In my experience, 'aboutEach' is in the category of 'nice try, but we
> didn't quite get it right'. 

Just because it's not "quite right" does not mean it isn't very
useful and shouldn't be retained.

Otherwise, on such criteria, we'd have to toss *alot* more stuff 
out of RDF, eh? ;-)

> I support a decision to drop it,
> *so long as* we give fair warning on www-rdf-interest and 
> elsewhere that
> this is our plan, and so long as we listen carefully to objections and
> concerns from the implementor community.

But the benefit is not to the implementor community. It is to
the content creators. I would be surprised if *any* implementor
would oppose a reduction in functionality requirements if they
are not themselves benefiting from them, eh? ;-)

> One way of doing this would be to offer an account of how 
> DAML+OIL, and
> hence potentially the new WebOnt WG's ontology language, offer similar
> facilities couched in terms of the graph data model.

Why should we wait for "future" specifications to provide
"someday" what RDF already provides now?!

I'm not saying aboutEach is perfect. I'm just saying it's very
useful, and being used.

Is it's limited implementation because causes a headache for 
implementors? Or simply that implementors haven't themselves
needed or used that feature?

I'd love to hear first hand, really...
 
> > </rdf:Description>
> >
> > <rdf:Description rdf:aboutEach="#gals">
> >    <rdf:type rdf:resource="x:Female"/>
> > </rdf:Description>
> >
> > This is also very useful for asserting "local"
> > classifications of resources for third-party
> > knowledge where you cannot touch the original
> > schemas.
> 
> I've not seen any deployment evidence of it being used in a
> distributed/hyertext style, though I agree this would've been 
> consistent
> with the original goal. 

Come on now! Be fair. Adoption of RDF has been as slow as
molases in January! Just because there aren't hundreds of
examples of use doesn't mean that they are not just around
the corner.

And I've informed you of one official use of this mechanism
(which I hope will be public knowledge soon), so I'm telling
you that there *are* uses of this, and I expect that there
will be lots more.

> There are other ways of achieving the 
> same end,
> either writing out the markup longhand 

Obviously. 

I explicitly said it's benefit was as a *convenience*.

> or using an ontology 
> language such
> as DAML+OIL.

As I said above. You shouldn't ask folks to stop what
they are already doing now in RDF because there *might*
be a better way to do it in the future.

That's not reasonable.
 
> > I know of one standards group which is in the
> > process of finalizing the decision to migrate
> > to RDF and is planning on using aboutEach a good
> > deal, though I'm not at liberty (yet) to say who
> > that is.
> 
> Perhaps you could alert them to the possibility that we are 
> considering
> dropping this feature of the language? Or, if/when we make such an
> 'advance warning' statement on www-rdf-interest, make sure it 
> gets passed
> on to them.

Rather, I will restate that I am opposed to the removal
of this feature, and will not likely be swayed unless 
implementors of RDF parsers can clearly explain to me
that it is a major pain in the rear end for them.

Cheers,

Patrick
Received on Friday, 16 November 2001 07:37:53 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 3 September 2003 09:42:43 EDT