Re: RDF MT: Default values for rdfs:domain and rdfs:range?

>Hi Jeen,
>
>Pat Hayes normally deals with MT questions, but he is only 
>intermittently around at the moment, so I'll risk a comment.

Pat is back, so will risk a comment|2

>
>At 14:27 28/03/2002 +0100, Jeen Broekstra wrote:
>
>>I am rather confused as to what exactly the domain/range of a
>>property should default to if no explicit domain/range is
>>specified,
>
>I don't think there is any notion of default.  In the absence of 
>domain/range constraints, then no constraints on the domain and 
>range are specified.

Right, exactly. If you want to think of the default as being in some 
sense the largest possible domain or range that could consistently be 
assumed (and hence the weakest constraint) then that would be 
rdfs:Resource for the domain, since the MT takes this to simply be 
the universe. Strictly speaking, the largest semantically possible 
range would be the union of rdfs:Resource with the set of literal 
values, but there is no name in the RDFS vocabulary which can be used 
to refer to that union.

However, the WG has now, I believe (Brian, correct me if this is out 
of date) decided that literals in RDF shall always denote character 
strings (ie the literal label itself, shorn of various XML taffeta), 
and so the rather fussy agnosticism exhibited by the current MT draft 
can be simplified by simply requiring that the universe always 
contains all such strings, which are assumed to be resources. This 
means that one can take rdfs:Resource to simply be a name for the 
universe, and so it can be used safely as a 'default' domain and 
range. The next draft of the MT will simply categorize urirefs and 
literals uniformly as 'names', and treat literals as a special kind 
of name that happens to have a fixed denotation, thereby simplifying 
the presentation considerably. (There was a time when we were 
contemplating extensions of RDF in which literals could denote the 
value domains of datatypes, rather than the strings themselves. Since 
these domains were open-ended, it was felt inappropriate to assume 
that they would all count as resources. But nobody seems to have any 
trouble with the idea that all character strings are resources. )

>Do you have a requirement for such defaults?
>
>>  and I was hoping your insights could clear it up for
>>me. I realize that the debate over what exactly constitutes a
>>resource and what a literal is far from over, but nevertheless I
>>hope that this practical issue can be resolved.
>>
>>In the case of rdfs:domain, it is relatively straightforward to
>>assume that if the domain is not specified, it is in fact
>>rdfs:Resource.
>
>Hmmm, it may be straightforward, but it is also potentially risky. 
>What if RDF were extended to allow literals as subjects.
>
>>  In the case of range the issue becomes cloudier
>>though, since legal values for objects include literals.
>>According to the RDF MT we should not confuse between literals
>>and their denotation, which I take to mean that literals can
>>conceivably be denoted by URIs. Furthermore, we assume that the
>>population of the class rdfs:Resource consists of anything that
>>could be denoted by a URI.
>
>I think we have been deliberately vague about the relationship 
>between literals and resources.  We have some charter contraints 
>that stop us getting into this fully.  So we are leaving things as 
>free as possible for a future WG.  You might think of this as some 
>deliberate underspecification.
>
>
>>This leads me to think that the range of a property also defaults
>>to rdfs:Resource, and that any use of that property with a
>>literal value is within that range. Is this correct?
>
>You may care to assume that.  I couldn't possibly comment.
>
>Sorry, English joke.  I won't say whether you are correct or not.  I 
>will say that RDF(S) makes no such statement.

Well, but maybe it will soon, right?

Pat

-- 
---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Tuesday, 2 April 2002 00:13:44 UTC