Review RDF: Concepts and Abstract Syntax (1rst part)

[Jerome posting on behalf of Jean-François who have trouble posting 
to the list]

RESOURCE DESCRIPTION FRAMEWORK (RDF):
CONCEPTS AND ABSTRACT SYNTAX

Sorry for being late for this review: I sent it yesterday, but gave one
of my mail address that is not registered in WebOnt: so it was not
accepted...

The main goal of this document is to define an abstract syntax for RDF
(core). It is thus a bridge between the two other RDF core normative
documents: "RDF Semantics" and "RDF/XML Syntax Specification".

This document is organized in the following way:

2 - MOTIVATION AND GOALS
3 - RDF CONCEPTS
4 - MEANING OF RDF
5 - XML CONTENT WITHIN AN RDF GRAPH
6 - ABSTRACT SYNTAX
7 - FRAGMENT IDENTIFIERS

My review will first point out some features of this document that could
cause problems to the OWL language(s). It is a basis of discussion for
the WebOnt working group. Second part of this review are personal
remarks/suggestions about the document, and as such should be directed
to the RDFCore working group.

A - RDF CONCEPTS AND ABSTRACT SYNTAX: IMPACT ON OWL
---------------------------------------------------

As far as I know, this document does not contain any problem with
respect to the abstract syntax of RDF, and examples of entailments
suggested in the document respect as well "RDF Semantics" as the
documents prepared in the WebOnt working group. Specifications for URIs
or datatypes encoding are out of the scope of my expertise.

The WebOnt working group should discuss about this document 4th section:
MEANING OF RDF. Excepted for the last subsection (4.5 Example), this
section is NORMATIVE. We can read that an RDF graph has formal semantics
(as defined in "RDF Semantics") as well as a social meaning. In 4.1: a
graph can be asserted (it makes claims about the 'real' world, so can be
used for inferences), or non asserted (meaning is "partly determined by
the circumstances [...]"). In 4.2: combining legal (e.g. saying
somewhere "the following graph is not true") and technical machinery
(e.g. an external media type to label information) to distinguish
asserted graphs from non asserted ones.
In 4.3: advice to use URIs defined by reputable organizations. In 4.4:
social meaning is applied to formally inferred consequences of a
document. In 4.5 (NON NORMATIVE): example showing who is responsible for
inference results whose social meaning can be considered insulting.

A1 - QUESTIONS TO THE WEBONT WORKING GROUP

1) ASSERTED/NON ASSERTED
Since proposed ways to distinguish between asserted and non asserted
forms are "opaque to logical reasoners", what confidence can we have in
RDF reasoners ? And since they are a subclass of OWL reasoners...

2)(SOCIAL/LEGAL) RESPONSIBILITY
The last example shows who is "socially responsible" for insulting
inferences using many RDF documents. The "social culprit" can be
identified for a RDF reasoner. This cannot be extended to OWL reasoners
(prime implicates ?). I do not know what can be the impact of this
notion of social/legal responsibility in such a normative document...

A2 - PROPOSAL(S)

I think that saying a graph can be asserted or non asserted (especially
in a normative section of a normative document)  weakens the semantics
of RDF (and by extension those of OWL), since nothing in the RDF syntax
provides a way to distinguish between those two cases. I see two
solutions:

	- remove this discussion, or at least weaken its impact. To
remove it, it is necessary to remove the 9th line of subsection 2.2: "A
basis for legally binding arguments", subsection 2.2.8, and all section
4 except the two first lines. To weaken it, change section 4 title (by
example "PUBLISHING RDF ON THE WEB"), make it non normative, and remove
all references to "legal bindings".

	- find a way to explicitely label a graph as non asserted in an
RDF graph itself. It should be possible using by example the reification
mechanism to describe a "nesting structure", as was done by example to
extend conceptual graphs. However, there should be still a lot of work
for a last call working draft...



Jean-François BAGET
EXMO Team
INRIA Rhône-Alpes

Received on Thursday, 13 February 2003 04:17:05 UTC