Re: review of XML in 10 points [was: AGENDA...]

(By the way, I do appreciate the attempt to get something about RDF,
ontologies, and W3C's Semantic Web activity into a widely-distributed
document.)


From: Bert Bos <bert@w3.org>
Subject: Re: review of XML in 10 points [was: AGENDA...]
Date: Fri, 7 Dec 2001 17:35:56 +0100

[...]

> I will mull over your text (thanks!) and see what changes it suggests.
> At first sight, though, it doesn't look as different as your use of
> "rubbish" had suggested. 

I think that you have severely underestimated my disagreements with the
paragraph in question.  

I'm going to make another try at stating my disagreements.


XML is the basis for RDF and the Semantic Web

XML provides 
an unambiguous syntax
	[What is unambiguous doing here?  Any syntax should be
	unambiguous.]
for W3C's RDF, the language for expressing metadata
	[RDF may be W3C's language for expressing metadata, but it is
	certainly not the only one, and not the only one that was designed
	for use on the web.] 
(in fact, for knowledge in general). 
	[This claim---that RDF can express knowledge in general---is the
	most dangerous kind of hype.  Readers who know little about the
	topic will take away the impression that expressing knowledge is a
	solved problem, which it certainly is not.  Readers who know more
	about the topic will take away the impression that W3C is clueless
	in this area, which I hope it is not.]
RDF is like hypertext elevated to the next level. Whereas hypertext links
pieces of text and leaves their relation vague, RDF can link anything and 
everything
	[This statement is again hype, but not nearly so dangerous as the
	hype above. RDF doesn't link anything to everything.  It can,
	perhaps, link formal objects can be considered to denote something
	to other formal objects.  ]
and assigns names to the relations: 
	[RDF also doesn't assign names to anything.  At best one could say
	that RDF can provide distinguishing tags on the relations and
	relate these tags to each other in interesting ways.]
'A is the price of B' can be a relation between an object and a sum of
money; 'A is heavier than B' can be the relation between two sumo
wrestlers; 'A is the cause of B' can be the relation between a shower and
your being wet.  
	[These examples severely overstate the reach of RDF in several
	ways.  RDF does not yet address typed data, needed for a decent
	handling of the first example.  RDF cannot do any causal reasoning,
	and cannot even represent states of affairs.]
To communicate knowledge, whether in XML/RDF or in plain English, both
people and machines need to agree on what words to use. 
	[Here there is a severe understatement.  Agreement on the words is
	insufficient---agreement on meanings is needed instead.  (Think of
	false cognates in natural laguages.)  Using ``words'' for machine
	communication is also a form of hype---machines do not yet have
	access to the richness of natural language, which is evoked by the
	use of the word ``word''.] 
A precisely defined set of words to describe a certain area of life
	[Again ``words'' is not an appropriate word to use here.  As well
	``life'' is one of those words that carries a gigantic halo of
	meaning along with it and should be avoided when talking about the
	current capabilities of machines.]
(from 'shopping' to 'mathematical logic') 
	[I'm not sure if ``mathematical logic'' qualifies as an ``area of
	life''.  In any case, RDF is totally inadequate for expressing
	any significant portion of mathematical logic.]
is called an 'ontology.' 
	[An ontology is not really a ``precisely defined set of words'' or
	even ``a precisely defined set of terms''.  However, for a
	paragraph of this sort the difference may not be worth worrying
	about.]
RDF, ontologies, and the representation of meaning so that computers can
help people do work are all topics of the Semantic Web Activity.
	[Agreed, at least to some reasonable approximation.]



In my view, the biggest change between the paragraph above and the
paragraph that I produced is that I drastically reduced both the scope
of and claims made for RDF.

Peter F. Patel-Schneider
Bell Labs Research

Received on Friday, 7 December 2001 12:25:14 UTC