Re: editorial tweak to OWL semantics doc

>From: pat hayes <phayes@ai.uwf.edu>
>Subject: Re: editorial tweak to OWL semantics doc
>Date: Mon, 16 Dec 2002 11:26:30 -0600
>
>>  >Hmm.
>>  >
>>  >Why shouldn't that go at the beginning of the RDFS-Compatible
>>  >Model-Theoretic Semantics section?  It appears to me that OWL/DL, when
>>  >written in N-triples,
>>
>>  Not N-triples; say RDF triples.
>
>Before I use the phrase ``RDF triple'' I would like to see a definition of
>them.  I looked in Concepts and Abstract Syntax and was highly
>underwhelmed.

Yeh, I don't much like the *way* that it is expressed there. Try the 
semantics document. One assumes three mutually disjoint sets: 
urirefs, literals (that set having a fairly complicated definition, 
but never mind) and blank nodes. Blank nodes can be anything that are 
clearly not urirefs or literals. RDF Triples are then defined in 
terms of those three sets, and RDF graphs are sets of triples. This 
is a mathematical structure, if you like; an abstract syntax. Now, 
any particular concrete syntax has to provide a lexical rendering of 
members of the three sets and of triples. Ntriples for example uses 
quote marks and various diacriticals to indicate literals, <angles> 
around urirefs, _:nodeIDs to indicate blank nodes, whitespace to 
separate them and dots to delineate triples. RDF/XML does it 
differently. Other surface syntaxes can use other forms, it doesn't 
matter a damn what they use as long as they are unambiguous about it.

But if 'RDF triples' sticks in your craw, just say 'in RDF' (or 'as 
an RDF graph') and leave it at that. Then its the RDF spec's problem 
to say what that means, right? All you need to know is that any RDF 
graph is a set of RDF triples, whatever the hell they are. You can go 
on using Ntriples notation to *describe* them, no problem there.

>  ``An RDF triple contains three components, ....''  Sure, but
>just *what* is an RDF triple?  Is it a triple, in which case why are the
>components named?

I don't follow that question or what point it is referring to.  The 
real answer is because those names were assigned by the old M&S and 
they are still useful. Do you see a problem with that? I don't.

>  Is it something else, like a data structure, in which
>case just what sort of data structure is it?

Its a structure. I guess you could call it an algebra, though that 
would be kind of pompous.

>
>I am unwilling to say that the (abstract) syntax of OWL contains elements
>that can be *anything*.  How are these sorts of thing supposed to be
>represented?

Any way that is unambiguously parsable into the RDF graph structure.

>
>So, unless there is a *decision* of the working group

This decision was taken well over a year ago and has been stated and 
re-stated many times since in emails and all extant drafts of the RDF 
documents and the closed-issues list, so it should hardly come as a 
surprise.

>, I am not going to
>use RDF triples until their definition is cleaned up or at least until I
>can understand just what is going on.

OWL/RDF is required to conform to the RDF standard and this is how 
the RDF standard is stated. If you can't bring yourself to use the 
terminology used in the prevailing standard documents then that seems 
to be a problem between you and your analyst. In the meantime, I 
would suggest that you ask the WG to get someone else do the editing.

Pat


-- 
---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola              			(850)202 4440   fax
FL 32501           				(850)291 0667    cell
phayes@ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
s.pam@ai.uwf.edu   for spam

Received on Monday, 16 December 2002 19:03:51 UTC