Re: RDF Issue rdfs-clarify-subClass-and-instance

At 07:13 AM 8/28/02 -0700, graham wideman wrote:
>Old: Class membership implies something about properties of instances
>New: Properties of instance imply class membership

Roughly, yes (to the 'New:').

I don't think there's been a recent change of understanding, even if the 
words themselves have changed recently.  RDF has always espoused the idea 
that "anyone can say anything about anything", against which background 
it's difficult to make sense of RDFS as constraints without nobbling its 
descriptive role.

So RDF Schema does not, of itself, limit what one can say about instances 
of classes mentioned in the schema -- e.g. someone else can always 
introduce additional statements using RDFS vocabulary that describe 
properties not conceived of by an original schema author.

But, this doesn't mean that this kind of use is completely dead:
Now, suppose I want to use RDFS for the mundane task of specifying 
structure and data of a few boring tables. I initially think that I can use 
rdfs:domain as part of my specification of which classes can have which 

One might want to use a schema in such a fashion to describe the vocabulary 
that is understood by a given RDF-processing application.  It then makes 
reasonable sense, given this schema and some arbitrary piece A of RDF, to 
ask the question:  "is 'A' something that can be understood by the 
application?", to which the answer may be "yes" or "no".  This is a kind of 
constraint test.  BUT, a "no" answer doesn't mean that A is invalid RDF, 
just that it's not fully understood by a given application.  Performing 
such a test involves making some logical assumptions that are not (and 
cannot be) expressed by the RDF alone -- in particular, assuming that all 
relevant information about a class (for the purpose at hand) has been made 

Looking further afield, I understand that DAML+OIL and work in progress of 
the OWL working group are building vocabularies that genuinely define 
restrictions corresponding to membership of a daml:Class concept, which can 
include requirements for certain properties to be present with particular 
kinds of value.  (I'm not sure if the ontology restrictions go as far as 
prohibiting appearance of properties not mentioned in a class specification.)


Graham Klyne

Received on Wednesday, 28 August 2002 11:29:41 UTC