Re: 2002-04-16 RDF Datatyping WD submitted for review by WG

On 2002-04-19 1:14, "ext Pat Hayes" <phayes@ai.uwf.edu> wrote:

>> On 2002-04-18 1:33, "ext Pat Hayes" <phayes@ai.uwf.edu> wrote:
>> 
>> 
>>>  Minor point. I know Im going back on what I said before, but now I
>>>  see all the examples (congrats on those, BTW) I find the closeness of
>>>  rdfd:datatype and rdfd:Datatype rather anxiety-producing.  (Does
>>>  anyone else agree?) I think we should either go back to rdfd:range or
>>>  something truly different like rdfd:typeCheckOnRange.
>> 
>> Would it help if we rather changed the name of the class to
>> rdfd:RDFDatatype?
> 
> Hmm, sounds like we are saying that it is one of *our* datatypes :-)

We are. The definition of an RDF Datatype is not identical to
that of an XML Schema simple datatype, even if it is a subset
of the latter.
 
> Actually I think that rdfd:Datatype is OK, its in line with
> rdfs:Class and rdfs:Property and so on.

That was my thinking.

Capitalized = class. Lowercase = property.

>> 
>> Thus, rdfd:datatype associates an rdfd:RDFDatatype
>> with a given property.
>> 
>> The reason why I moved away from using rdfd:range is that it
>> is not the rdfd:range property which is constraining anything,
>> it is the semantics of the datatype itself, which is completely
>> opaque to RDF. All that the rdfd:range/datatype property does
>> is associate a datatype with a property, so that all datatyping
>> idioms used with that property are interpreted in terms of that
>> datatype.
> 
> Right, but it does it in a particular way, see below.
> 
>> 
>> The datatype itself constrains the literals to the members of
>> its lexical space and the bnodes to the members of its value
>> space based simply on the presence or absence of a lexical
>> to value mapping from the literal/lexical form to the value.
>> If there is none, then either the literal/lexical form or
>> value (or both) are invalid.
>> 
>> Thus, there is no range like semantics asserted by the
>> rdfd:range/datatype property itself in the same manner
>> as the rdfs:range property. So I wouldn't want to revert
>> to rdfd:range.
> 
> Right, I take your point and agree. On the other hand there is a link
> to rdfs:range in that the datatyping constraints are applied to the
> object of the property rather than the subject, so the 'pointed end'
> association is similar to rdfs:range (instead of rdfs:domain). Its
> not a very logical connection, but I bet users will find it a useful
> association. That is why I rather liked the 'drange' or 'dcrange'
> idioms, which had a kind of range-ish flavor but didnt actually say
> 'range'.
> 
> How about rdfd:dtype ??

Hmmmm... maybe. Though there is the potential confusion with
rdf:type -- and rdfd:dtype would not be anything like rdf:type.

> I really just want it to be more visually
> distinct from the class name: it's too easy to make typos at present.

I do appreciate that point.

Let's continue to think of possible alternatives. I'd like, though,
to avoid any confusion with any existing RDF property since the
semantics of the datatype association is quite unique.

Patrick

--
               
Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email: patrick.stickler@nokia.com

Received on Friday, 19 April 2002 04:10:07 UTC