Re: completion of action: 2001-07-27#2 (long)

pat hayes wrote:
[...]
> 
> >Similarly, in the mathematical concept of a tuple, e.g. (x,y,z) can
> >the components be physical objects, even though the tuple itself is
> >an abstract thing?
> 
> Sure, an *arbitrary* mathematical triple. But if you go on to talk
> about these triples as constituting a *language* - one that can be
> parsed, has a grammar, and so on - then you are no longer understood
> to be talking about arbitrary triples, but those containing
> language-like entities, usually symbols of some kind. The mathematics
> you were using to talk about the things in the triples is itself a
> language, of course, so when you write
> (x,y,z)
> you are both using and mentioning a triple. The one you are using is
> linguistic in nature, and might reasonably be claimed to be the
> abstract syntax of the lexical string you wrote (a string of 7
> characters which parses into an abstract entity which is a triple
> containing three one-character lexical entities); but if you were
> using 'x', 'y' and 'z' to refer to, say, backhoes, then the triple
> consisting of three backhoes is most definitely not what is usually
> thought of as an expression in a language. BNF is about *character*
> strings, not sets of sets of backhoes, and the labels of nodes in an
> RDF graph are URIs, not backhoes, right? Things that can be sent
> along wires, using transfer protocols.

M&S makes a distinction between a statement and a triple.

My interpretation of it, is that a statement is an abstract mathematical
tuple containing three components, the subject - a resource, the predicate
- a property and the object - a resource or literal.

A triple is a representation of a statement, e.g. in a language, or in 
the memory of a computer or as signals on a wire.

This is how I make sense of what M&S says about statements containing
the resources themselves and

  http://lists.w3.org/Archives/Public/www-archive/2001Jun/att-0021/00-part#291

 Triple
     A representation of a statement used by RDF, consisting of just the
     property, the resource identifier, and the property value in that order.

That said, the glossary definition of a statement:

  http://lists.w3.org/Archives/Public/www-archive/2001Jun/att-0021/00-part#289

seems inconsistent with that defined in section 5, and I tend to give section
5 more weight.

> 
> It may be that RDF is predicated on a much more revolutionary idea
> about the nature of language than I had previously realized. Do you
> really want to claim that an RDF graph consists of triples of
> *anything*, even of physical things?

M&S's treatment of statements as being composed of the resources themselves
has led to all sorts of problems and overly long discussions on RDF interest
and elsewhere, in part because nobody seems to be really sure what a resource
is.

That was why I was keen that we move away from the set oriented model as
defined in M&S to what we have now where we have a language, the RDF Graph,
and an interpretation function that maps URI's to pumpkins.  This avoids
so much of that mess.  Its a great step forward.  Joy unbridled and all
that ...

Now when we get into reification, some of this mess reappears. I'm not 
trying to defend M&S's position.  But it did seem right to point out
what M&S says on the subject. Maybe I shouldn't have brought it up right now. 

I suggest we continue with our plan, punt reification till later and hammer
a stake in ground with what we have now.

Brian

Received on Wednesday, 5 September 2001 06:45:34 UTC