Re: DAML+OIL, RDF, XML & WebOnt

From: "Jeremy Carroll" <jjc@hplb.hpl.hp.com>
Subject: DAML+OIL, RDF, XML & WebOnt
Date: Fri, 7 Dec 2001 11:20:54 -0000

> Summary: DAML+OIL is built on RDF, not XML.

Yes, at least sort of.   This was more the case originally than it is now
(because RDF has changed).

DAML+OIL is built on top of RDF syntax.  It even tries to stay close to the
intent of RDF.  However, at the time DAML+OIL was designed RDF did not have
a model theory, and DAML+OIL is not upward compatible with the new RDF
model theory.

The big problem is that there are a large number of assumptions built into
various parts of RDF.  Some of these assumptions are acceptable within RDF
and within many logical formalisms, some of them are acceptable within RDF
but not within more-expressive logical formalisms, and some of them are
just not acceptable at all.  (Some of these last are being addressed by the
RDF Core Working Group.)

Here is a partial list of the assumptions.   Note that not all designers of
RDF may agree that RDF makes all of these assumptions, but they each are
explicitly in one or the other of the RDF specifications or can easily be
read into the specifications.

1/ Data models are adequate for providing meaning for logical formalisms.
2/ Triples are adequate for representing all knowledge.
3/ Syntactic reification is adequate for representing knowledge about knowledge.
4/ Syntactic reification is adequate for extending a logical formalism.
5/ Theory and metatheory can be written and used in the same way.
6/ Formal specifications are not necessary.
7/ URIs with fragments are general identifiers.
8/ String literals suffice to represent all data.
9/ Containers do not need a theory of equality.
10/ Extensions that require extra machinery do not need to describe how
    that machinery is to work.

The designers of DAML+OIL had to deal with these assumptions.  I think that
we did a reasonable job, but we may not have been sufficiently forthright
as to which of these assumptions we had to violate.  (For the record, I
think that we violated all of them except for the seventh, which we didn't
think much about, and the nineth, which we didn't address at all.)


> Conclusion from below:
> [[[
> DAML+OIL assumes full RDF processing with a compliant RDF parser.
> 
> A DAML ontology written in N3 or N-triple is still a DAML ontology.
> 
> This is in contrast with other RDF dialects such as RSS [7] or CC/PP [8]
> where many implementations assume a rigid XML document format, with no
> variation from the examples in the specs.
> ]]]


[...]


peter

Received on Friday, 7 December 2001 09:06:40 UTC