RDF LCWGs comments: RDF core, database example, soc-entailment and typo

Dear all,

I have few remarks concerning the RDF specifications (LCWGs).

Outline:

#1. RDF-CONSEPTS: The term RDF core is used vaguely (e.g. in sect 1).
#2. RDF-CONSEPTS: Example in 3.5 fails to model an actual database?
#3. RDF-CONSEPTS: Characterisation on 4.4 about the social meaning of RDF
entailments is too strong to be acceptable.
#4. RDF-SEMANTICS: A typo in 3.1 in the term rdfV?

Here are the full comments:

#1.) In document Resource Description Framework (RDF):
Concepts and Abstract Syntax
(http://www.w3.org/TR/2003/WD-rdf-concepts-20030123/) the introduction
uses the term "RDF core" rather vaguely, e.g. by saying that:

"The framework is designed so that vocabularies can be layered on top of a
core. The RDF core and RDF vocabulary definition (RDF schema) languages
[RDF-VOCABULARY] are the first such vocabularies."

... so RDF core is a vocabulary build on top of a core?

Perhaps the intended meaning of the concept "RDF core" (the WG, the
fundamental set of RDF specs, or something else?) should be spelled out
better since the term "RDF core" will be probably widely used in the
discussions about RDF hereafter?

#2.) In document Resource Description Framework (RDF):
Concepts and Abstract Syntax
(http://www.w3.org/TR/2003/WD-rdf-concepts-20030123/), the subsection 3.5.
"Representation of Simple Facts" provides an example of encoding a
database row as a set of RDF statements.

The example is nice but perhaps a bit misleading since databases come with
the basic assumption of unambiguous data. If I'm not mistaken, the
semantics of the RDF core specifications (RDF Schemas in particular) do
not provide a validation mechanism preventing the occurrence of multiple
values for the cells when the database is modelled as in the example. For
instance, someone might simply write two sentences

_:x http://.../city "Bedform"
_:x http://.../city "Berlin"

which effectively would brake the idea of a database. (For predicates this
of course is no problem since they are effectively relations.)

#3.) In document Resource Description Framework (RDF):
Concepts and Abstract Syntax
(http://www.w3.org/TR/2003/WD-rdf-concepts-20030123/), the subsection 4.4.
"Interaction Between Social and Formal Meaning"
(http://www.w3.org/TR/2003/WD-rdf-concepts-20030123/#section-Interaction)
reads:

"The meaning of an RDF document includes the social meaning, the formal
meaning, and the social meaning of the formal entailments. The assertion
of an RDF graph G, when G logically entails G', includes the implicit
assertion of G'. The implied assertion of G' should be interpreted using
the same social conventions that are reasonably used to interpret the
assertion of G."

This sounds like a rather strong normative characterisation (perhaps too
strong).

First, even if I believed that (human or artificial) agents would mutually
agree a mechanism for making global vocabulary entailments, I don't think
that agents in general are capable (or willing!) of stating only RDF
sentences whose entailments they fully agree (or can come feasibly up with
and interpret). In addition, by introducing a vocabulary entailment
including negation, it's easy to end up with unintentional entailments
(effective anything). More reasonable would be saying that agents must
either agree on the demonstrated entailments, refine their assertions, or
choose a different interpretation theory altogether.

Second, in practical situations, there are typically several possibilities
for the selection of the RDF graph, only some of which an agent is either
capable or willing to consider. For instance, I might only have access to
graphs G1, G2, and G3, but willing only to accept the assertions in G1 and
G3. Thus if G1 asserts {A->B, E->C}, G3 asserts {C->D}, and even if G2
asserts {B->D}, I would accept only {A->B, E-C, C->D} but not {A->C}.
However, there might be a graph G4 (that I, e.g., don't know about) which
entails {B->E}, using assertions and inference mechanism that I would
accept.

From mathematical point of view, the result would be the same (accept G2
or not, by accepting G1, G3, and G4 one would entail A->C), but from the
metamathematical point of view, quite different (different proof)--I might
argue the same thing as the next guy but with different arguments. In the
social context this is extremely important (consider, e.g., the system of
law). Thus the entailment path itself should be on focus as well as the
result.

A trivial example to emphasise this point:
G1: {A->C}
G2: {A->B, B->C}
Socially accepting G1 alone might be harder that its "proof", G2 (which
logically does entail G1). (Consider: "JohnSmith is a clown" versus
"JohnSmith acts foolishly" and "foolishly acting people are clowns" thus
"JohnSmith is a clown".)

Perhaps RDF should introduce new concepts, for denoting the particular
graph an agent is using (capable and willing) while doing deductions and
interpretations, and the notion of a proof (i.e., some assertions are
"more valuable" than others since there is a widely accepted proof for
them)?

#4.) In document RDF Semantics
(http://www.w3.org/TR/2003/WD-rdf-mt-20030123/), the first paragraph of
the subsection reads:

"RDF imposes some extra semantic conditions on the following (rather
small) vocabulary, which we will call rdfRV:"

and goes on introducing the RDF Vocabulary in a table.

It seems that the name "rdfRV" is a typo which should be replaces with the
name "rdfV"?


In general, I very happy to see RDF taking shape, thanks!

Best regards,

--Ossi

--
Ossi Nykänen                              Tel   +358 3 3115 3544
Tampere University of Technology          Fax   +358 3 3115 3549
DMI / W3C Finnish Office                  Email ossi@w3.org
P.O.Box 553, FIN-33101 Tampere, Finland   Web   www.w3c.tut.fi

Received on Friday, 31 January 2003 06:21:41 UTC