W3C home > Mailing lists > Public > www-rdf-logic@w3.org > November 2000

What is an Ontology? [was: daml-ont]

From: Dan Connolly <connolly@w3.org>
Date: Wed, 01 Nov 2000 11:02:43 -0600
Message-ID: <3A004CB3.1F700934@w3.org>
To: "McBride, Brian" <bwm@hplb.hpl.hp.com>
CC: www-rdf-logic@w3.org
"McBride, Brian" wrote:
> 
> In the comment property defining Ontology, it states that an
> ontology is a document describing a vocabulary ...
> 
> Is the ontology the document, or is the ontology an abstract
> entity, and the document one of many possible
> representations of it?

There are (at least) two senses of the term 'document'
in web literature:

"document 
      1.aka resource; aka Node; See also: visit 
      2.aka page, frame, card 
      3.a bit of data 
            in [SGML86] 
            in [HTML95], 3.2, 4.0 "
	-- http://www.w3.org/Architecture/Terms#document

In the comment property on Ontology, I was using
document in the first sense, i.e. an "abstract
entity" that may have many possible representations,
rather than in sense 3, a particular byte-sequence.
(hmm... "visit" really belong with sense 2,
i.e. a compound document with embedded images
and such.)

Aside from traditional dublin-core style author/date/title
metadata, I expect Ontology resources
to be relevant in the case of assertion by
reference. cf

Assertion by reference [was: Comments on Annotated DAML...]
Dan Connolly (Wed, Oct 11 2000) 
http://lists.w3.org/Archives/Public/www-rdf-logic/2000Oct/0005.html

My model is based on a sort of modal-logic
approach to integrating RDF statements into the
real world of communication protocols; consider
a property
	entails: Message -> Statement	
	for entails(m, s) read: the message m entails
	the statement s

and
	about: Message -> Resource
	for about(m, r) read: message m is about r;
	e.g. when you fetch http://example.org/page1,
	the HTTP GET message you send, and its reply,
	are about http://example.org/page1

These allow me to state the axioms I have in
mind for daml:imports ; namely:

	entails(?xReply, `daml:imports(?x, ?y)')
	/\ about(?xReply, ?x)
	/\ about(?yReply, ?y)
	/\ entails(?yReply, ?aStatement)
	=> entails(?xReply, ?aStatement)

where `p(s, o)' denotes reification/quoting
[and is kinda hairy in ways that I haven't
fully debugged].


>  One might wish to make statements
> about the abstract entity which are not true of the document
> and statements about a document which are not true of the
> abstract entity.

Indeed... the document(sense3) that is the content
of ?xReply above isn't (necessarily) the same thing
as ?x.

>  These should therefore have different URI's.

Right... if you want to refer to the document(sense3)
that is the content of ?xReply, you may need
to use a URI that's different from the one
you use to refer to ?x.

> In the light of that, should the fomulaic:
> 
>   <Ontology about="">
> 
> have an absolute URI of the ontology between the "'s as the
> element is describing the ontology, not the document.

I disagee; about="" is a very convenient way to refer
to that-wich-this-message-is-about.

> [I am in general supportive of Sergey's stylistic suggestion
>  that in general, URI's in about attribute values should be
>  absolute - it is rather irritating to cut and paste a self
>  contained chunk of RDF and find the semantics have changed.
>  Also, if a mirror service stores an exact copy of an RDF
>  document, to have the copy have different semantics to
>  the original seems unhelpful.]

It's entirely feasible to mirror some content without
replublishing it at a different address. For example,
the W3C web site is mirrored at a dozen or so
locations in IP-space (and physical space) but all
the mirrors are at the same address in Web space:
http://www.w3.org/ .

Republishing stuff at a different address is
another matter entirely, and folks shouldn't
get the impression that it can be done without
complications in the general case: relative
URI references are a small technical matter
compared to access control, metadata
about copyright/licensing, digital signatures,
etc.

Style is a matter of taste and experience.
My experience leads me to favor relative URI
references.



> There are a couple of minor nits in the example ontology:
[I'll address these separately.]


-- 
bind default <http://www.w3.org/People/Connolly/kb>
<mailto:connolly@w3.org> is mailbox of 
  [a Person; called "Dan Connolly";
  affiliation [ a Consortium; called "W3C";
	     homePage <http://www.w3.org> ];
  homePage <http://www.w3.org/People/Connolly/>;
 ]
Received on Wednesday, 1 November 2000 12:03:59 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 2 March 2016 11:10:32 UTC