W3C home > Mailing lists > Public > www-rdf-logic@w3.org > May 2001

Re: metadata vs. meta-(meta)-data

From: Dan Connolly <connolly@w3.org>
Date: Tue, 22 May 2001 11:46:27 -0500
Message-ID: <3B0A97E3.D2695EE2@w3.org>
To: Larry Masinter <LMM@acm.org>
CC: www-rdf-logic@w3.org
Larry Masinter wrote:
> 
> I'm hoping that RDF can be modified in a way that
> distinguishes between the use of RDF to make assertions
> about actual web resources (data content) and its use
> to make assertions about things that are described or
> referenced by things on the net.
> 
> The use of "http" URIs as unique identifiers for things
> that are not resources backed by HTTP web servers is
> a serious confusion of levels.

Er... it's a straightfoward contradiction, no?
http://mycollege.edu/students/Amy is definitely
an HTTP thingy; if something tries to say that
it's not, that's a contradiction between that
something and the (axioms taken from the) HTTP/URI specs.

> The model for the sentence
> 
>  The students in course 6.001 are Amy, Tim, John, Mary, and Sue.
>  is written in RDF/XML as
> 
> <rdf:RDF>
>   <rdf:Description about="http://mycollege.edu/courses/6.001">
>     <s:students>
>       <rdf:Bag>
>         <rdf:li resource="http://mycollege.edu/students/Amy"/>
>         <rdf:li resource="http://mycollege.edu/students/Tim"/>
>         <rdf:li resource="http://mycollege.edu/students/John"/>
>         <rdf:li resource="http://mycollege.edu/students/Mary"/>
>         <rdf:li resource="http://mycollege.edu/students/Sue"/>
>       </rdf:Bag>
>     </s:students>
>   </rdf:Description>
> </rdf:RDF>
> 
> This example shows the confusion -- the description is not
> about the web page at http://mycollege.edu/courses/6.001, but
> it's about the course which is described by the web page at
> http://mycollege.edu/courses/6.001.

It's less than ideal, yes, but I can see several ways to
make sense of it:

(a) in the schema for s:students, it could say

	This property relates a web page about a class
	with a collection of web pages about students in the class.

(b) if the schema syas

	this property relates a class with a collection
	of students in the class

Then the above RDF blurb says that the class and the
web page are the same thing. In the absense of
axioms/constraints to the contrary, that makes
logical sense, no?

i.e. there's no logical problem with saying that
the integer 2 and the pencil I'm holding in
my hand are identical, are there?

I'd recommend against it as a modelling practice,
but I don't see a problem with RDF per se.

> However, we may also want
> to make assertions about the web page: when it was written,
> who wrote it, when it was last updated, where it is archived,
> etc.

What's to prevent them from doing so?

The only issue is that if they take approach (b) where
they equate the HTTP resource with the class, they can't
say that X wrote the class and Y wrote the page
without also saying that X wrote the page and
Y wrote the class.

> The web page and the thing the web page is about are at
> different levels of quoting, but these levels are confounded
> in RDF today.

Hmm... to say that a whole document denotes something is
interesting; I wouldn't have looked at that as quoting
so much as a normal relationship, sorta like dublin
core Subject.

I dont' see anythingabout RDF that forces folks
to confuse levels. (There's nothing that
prevents them either, as demonstrated above.)

> Adding a "#" at the end of the Description's about attribute's
> URI doesn't really help, since it resolves into the structural
> granuarity of the web page rather than into the next level of
> reference.

I don't think that http://www.w3.org/People/Connolly/#dan
necessarily denotes a structural part of a document.
What's wrong with saying that it denotes me?

i.e. http://example/#FRAG denotes whatever
#FRAG denotes in messages received in replies
to GET requests to http://example/.

But maybe this is a different sort of reference
than the one used in <a href="...">. Hmm...


> I can think of several ways of fixing this, but they're all somewhat
> unpleasant:
> 
> - leave RDF as is, add another URI scheme that means
>   'the thing described by this URI'
>   <rdf:description about="ttdb:http://mycollege.edu/courses/6.001">
>   (ttdb - the thing described by )

That doesn't appeal to me at all. I'd rather use a
relationship...

	<rdf:Description rdf:about="http://mycollege.edu/courses/6.001">
	  <x:describes>
	    <s:Course>
              <s:stundents>...</s:students>
            </s:Course>
          </x:describes>
        </rdf:Description>

(I've wondered if dc:subject should be used this way;
i.e. does the Dublin Core community expect dc:subject
to be used to relate a photo to a person that
appears in a photo? or does it always relate
the photo to subject keywords?)


> - leave RDF as is, assume ttdb, and add a level of quoting if
>   you want to use RDF for metadata, e.g.,
>   <rdf:description about="data:text/uri,http://mycollege.edu/courses/6.001">

Again, why make an expensive new URI scheme when you
can express it with a cheap new RDF property?

> - Define that each relationship should be specific about its level
>   of indirection, e.g., "dublin core RDF relations are about the
>   URI as a web resource, but other kinds of RDF assertions might
>   really about the thing described by the web resource rather than
>   the web resource itself".

I wouln't call it "level of indirection," but yes,
the specification of each RDF property needs
to be clear about what its domain and range is,
ala (a) above.

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
Received on Tuesday, 22 May 2001 12:46:32 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 2 March 2016 11:10:35 UTC