RDF isn't so hot

At the risk of throwing in a wet towel on a flame fest etc. etc. I feel it 
is important to raise a few warnings about RDF.

 From a number of different points of view, RDF has serious issues:

1. Software Engineering
   RDF is `aggressively' untyped; in some ways it is even worse than XML in 
this regard which at least has DTDs and XML schemas to assist the process.
  RDF is untypeable (sic) and proud of it. This maximizes the impedance gap 
between RDF and regular programming languages.

2. Logical expressiveness
   RDF is a very simple language, propositional in character, when viewed 
as a language for expressing knowledge. This puts a serious dent in its 
utility. DAML `solves' this by imposing a somewhat artificial layering on 
top of RDF -- to the point where DAML is both crippled by its foundations 
and in fact pretty distant from them. The logical technique used in the 
DAML semantics seems (to this person) a little dubious.

3. Semantics
   Taken as a weak KR language (which is its purpose) RDF appears to be 
higher order. Since it is possible to state, in RDF:
   likes = hates

   Pat Hayes has developed a semantics for RDF that skirts this problem but 
hey: we have a weak, untyped language that needs some sophisticated logic 
to get a reasonable semantics. That sounds promising!

Let me add one important point: the MOTIVATION for RDF, and DAML/OIL for 
that matter, is spot on. This complaint is about the technology used to 
solve the problems.

Remember, there is nothing wrong with problem solving; but all technology 
solutions show up on the negative side of the equation -- the RISK side.

Frank

Received on Monday, 12 August 2002 12:36:08 UTC