W3C home > Mailing lists > Public > www-rdf-interest@w3.org > April 2002

RE: Disambiguation; keeping the "U" in "URI" (& Documents, Cars, Hills, and Valleys)

From: Manos Batsis <m.batsis@bsnet.gr>
Date: Thu, 25 Apr 2002 14:30:24 +0300
Message-ID: <E657D8576967CF448D6AF22CB42DD2690FF264@ermhs.Athens.BrokerSystems.gr>
To: "Danny Ayers" <danny666@virgilio.it>
Cc: <www-rdf-interest@w3.org>

Hi Danny, 

> -----Original Message-----
> From: Danny Ayers [mailto:danny666@virgilio.it] 

> I think the key may lie in Thomas' remark earlier about 'it 
> only pushing it
> up a level'. Up a level we have a lot more power.

Yes but one level up always hides some things, making it inconvenient to
step down and do the dirty work; Generally speaking, this may become an
akward way of doing things.

> What if we consider the URI to represent a *set* then the assertions :
> 
> A http://www.markbaker.ca/index.html has long hair
> B http://www.markbaker.ca/index.html is hosted in Florida
> 
> can be made 'in the wild', in that A & B refer to a different 
> element in the
> set http://www.markbaker.ca/index.html

I would say that each property refers to a member in the set of what
http://www.markbaker.ca/index.html may represent under any context.

> If we wish to reason with such statements, then locally we 
> can pull out the
> element of the set of interest, and give it a local unique 
> identifier (if
> necessary).

What is the mechanism you have in mind?

> If we want to harvest assertions about  
> http://www.markbaker.ca/index.html
> the person, then we look 
> for those statements containing this URI, but only
> with predicates that apply to or on a person.

Looks great at first glance but looking at the rdfs:domain to determine
what the URI represents in any case is not always accurate. Properties
may overlap between one or more candidates since they are inherited.
Also, your mechanism is based on rdf:Property declarations that are
global or uniform, which may bring back the uniqueness problem from the
dead. Makes the whole use of URIs unreasonable.
I like the approach but it may need some work.

> Now shoot me down in flames...

No flames from this side... Just some considerations.

Let's make our requirements clear for a second, make sure what we are
after.

From what I understand (and please think of this without my poor
terminology), we all see a URI as a representative of a set of abstract
entities, including itself as a document (the default?).
Under the above "definition", what we may need is to enhance RDF with an
abstract data structure (a class) where any property addressing the
resource/class is aimed at *one* member of that class. This can simply
be done with a list of what the URI may represent using a new property.
But... This property will have URIs as values so what have we done? A
full circle? Urgh..

Cheers,

Manos
Received on Thursday, 25 April 2002 10:02:46 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:51:53 GMT