Re: I have a trouble with The RDF Model

Some thoughts on a RDF data model and 
schema evolution ::

When we view reification as a relationship, we
can get a directed graph representing entities and 
relationships.  In various domains, the design
approach is top-down (starting with high level
entities).

Here also entities are sentence types with 
one or more properties.  

       E1(A,B,C) -------------->  E2(P,Q,R)
         (1/1)     is_reified        (1/N)

One can get inheritance and polymorphism by 
assigning restrictions in the relationships 
(Entities can be implicit or explicit). 

However, unlike other areas of schema design, RDF
schema design could be nontrivial. Most
other domains implement schema evolution in a limited
way.  Schema evolves when an entity or
an attribute or a relationship gets added,
deleted or modified.  Because World Wide Web evolves
in an unpredictable way, RDF schema should also
evolve to adjust to such changes. A graph G(S1, S2)
with two sentences can evolve into a graph with
two sub-graphs G(G1, G2). An entity E(A, B, C) with 
simple properties can evolve into E(A, G, C) where 
G is a graph. A relationship between E1 and E2 can
change to a three-way relationship over E1, G and E3
where G is a graph.

These nontrivial situations cannot be handled unless
we also preserve functional dependencies 
over properties and adopt a bottom-up approach in
conjunction with top-down
approach. I think that one needs to exploit here all
techniques in schema design (both formal and 
informal).

-- ssarkar@ayushnet.com

Received on Friday, 1 December 2000 12:10:39 UTC