Re: [xml-dev] URIs are simply names was:Re:[xml-dev]"Abstract"URIs

On 2002-02-20 17:40, "ext Manos Batsis" <> wrote:

> I guess that if people are interested in this, conversation should take
> place in RDF-interest.
> | -----Original Message-----
> | From: Patrick Stickler []
> | On 2002-02-19 16:59, "ext Manos Batsis" <> wrote:
> | > Take RDF for example, (AFAIK and even worse, last time I
> | checked) where
> | > literals cannot be subjects in a sentence (triple). This
> | essentially has
> | > to do with the RDF model; a literal there is just that; what if the
> | > system was able to uniquely identify this literal and refer
> | to it with a
> | > *unique* URI; all implemented features applicable to
> | resources would be
> | > available for literals as well.
> | 
> | I don't see the point. Do you wish to talk about an occurrence
> | of a literal? A concrete example would help.
> Sorry about that; in short, I would like to see RDF having [resource] at
> the top of the object hierarchy, where literals are also treated as
> subclasses of [resource] and  of some primitive type probably taken by
> the xsd namespace. This is common:

One word of fore-warning, the RDF Core WG is in the heat of
deliberation over datatyping, and so these issues are, and
have been, under discussion for some time now.

See the RDF Core discussion list archives for all the gory

> :manos xx:age "22".
> How about
> "22" aa:typeOf xsd:int
> Of cource the above idea crashes when we add this to the picture
> :mySweater xx:label "22"
> "22" aa:typeOf xsd:string
> We have a conflict.

If literals are tidy (i.e. all literal nodes with the
same string-equal label are merged, yes.

If each literal node is unique, then there is no problem.

It all boils down to whether you maintain the unique
context of each literal occurrence.

The present consensus in RDF core is that literals will
be tidy and thus context is not born by the literal node
itself in the graph.

> Ok, from scratch. How about this:
> :mySweater xx:label xsd.string:22

This is a TDL URV, a typed data literal uniform resource value.


This is a great approach. Some folks don't like it for
some unknown reason...

> There are no conflicts; for example
> :foo1 xx:size
> :foo2 xx:heightInCm
> :mySweater xx:label xsd.string:22
> I'm trying to deal with primitives as unique resources because I see
> them as such; the difference is they are not addressable as URLs, since
> a string "ab" is unique, as any string or number or any other primitive.
> So above, xx:size and xx:heightInCm have the same resource as an object;
> that resource is the unique "literal of type int:22".

Literal strings are not globally unique resources. They can't be.

In a closed system, one might ascribe them consistent meaning, but
in a context of large syndication of knowledge from arbitrary sources,
this doesn't work.

A literal must have context to have consistent interpretation.

Much of the effort going into RDF datatyping is finding a way
to express literals-in-context which consistently and unambiguously
denote actual values. RDF can't provide explicit representation of
datatype values since RDF has no native datatypes (and shouldn't).
The best we can achieve is a consistent and unambiguous denotation
of a value.

> Would something like the above ban literals as used today in the RDF
> model, while introducing the primitive_typing mechanism as most
> anticipate it?

That question is currently being debated. My view is that there
is a simple way to avoid such a conflict. That view is under
consideration. There are other views. Though I think that the WG
has converged pretty much on all of the key aspects of the matter.

We'll see (and real soon now, we're supposed to be done already ;-)



Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email:

Received on Wednesday, 20 February 2002 11:40:50 UTC