Re: Datatyping desiderata, take 2

At 03:18 PM 1/17/02 +0200, Patrick Stickler wrote:
> > Also, I lost the point of the list of idiom examples: idiom B and idiom
> > P are shown to be the same (modulo prefix). Have I missed something?
>
>They are almost, but not quite the same. In the P idiom, the
>rdfs:range value is a URI that denotes the entire datatype,
>whereas in the B idiom, the value is a URI that only denotes
>the lexical space of a datatype. In the S proposal, the
>type URI is given a suffix '.lex' (e.g. ex:date.lex). Is there
>perhaps an error in the example for B? Should it be 'ex:date.lex'
>rather than 'ex:date'? The same question comes to mind for the
>example for A -- should it be 'ex:date.map' rather than 'ex:date'?

Currently, they are exactly the same, because in presenting the idioms I'm 
trying to avoid saying anything about the mechanisms of interpretation 
(after all, that's what we're trying to choose, right?)

As presented, the idioms are simply forms of RDF-graph that all have the 
same intended meaning:  to express the idea that Jenny was born on 15th day 
of the 7th month of year 2001.  These idioms do not attempt to say anything 
about the mechanisms of interpretation that get us from the original graph 
to the intended meaning (just as most of the RDF users who have used these 
idioms to date have no view of how the intened meaning is formalized).

Thus, two idioms which have the same graph form (modulo prefix renaming) 
and the same intended meaning are, ipso facto, identical idioms for the 
purpose of this discussion.

As I said previously, I've maintained them separately so far because 
subsequent discussions may show that extended versions of these idioms 
don't turn out to be identical.

Some confusion may arise because I've named two of the idioms (P and D) for 
the schemes for which they were presented as examples.  That was probably a 
bad idea.  I was simply trying to give some clue as to the provenance of 
the idiom, and not to say anything about its mechanisms of 
interpretation.  I think I'll just call them A, B, C and D.

>It occurs to me that perhaps the common desiderada should not
>attempt to summarize the different idioms, as they tend to depend
>on an understanding of each particular datatyping scheme, but
>that it should be left to each proposal to describe the idioms
>employed/recommended, with such descriptions given in terms of
>the semantics of each scheme.

My point here is a backwards-compatibility argument:  the idioms are 
claimed or suggested to be in current use, and absent any particular 
mechanism of interpretation.

>...  I.e., it is only by understanding
>the S scheme that one knows that ex:date in idiom A refers to
>the mapping (not the lexical space or value space) of the
>datatype and in idiom B it refers to the lexical space, etc.

Indeed, the scheme explains how one arrives at the intended 
interpretation.  That is the goal of our debate, I think.  The presentation 
of idioms is intended to be agnostic of such matters.  The selection of a 
datatyping scheme will
(a) tell us which of the idioms can be supported, and
(b) explain how the intended interpretation is obtained for each supported 
idiom.

#g


--------------------------
        __
       /\ \    Graham Klyne
      /  \ \   (GK@ACM.ORG)
     / /\ \ \
    / / /\ \ \
   / / /__\_\ \
  / / /________\
  \/___________/

Received on Friday, 18 January 2002 08:57:33 UTC