Re: What do the ontologists want

From: Sergey Melnik <melnik@db.stanford.edu>
Subject: Re: What do the ontologists want
Date: Thu, 17 May 2001 14:07:24 -0700

> "Peter F. Patel-Schneider" wrote:
> 
> > [...]
> > I think that Pat has precisely characterised the situation.
> > There have been numerous posts on this group extolling the virtues of
> > reification.  Many times I think that I am listening to late-night
> > infomercial:
> > 
> >         Reification---it chops, it slices, it dices.  No language should be
> >         without it.  No language needs more than it.  Get yours now.  Easy
> >         payment terms of just endless hours of frustration from now to
> >         eternity.
> > 
> >         [Only partly in jest.]
> 
> Despite all due frustration, reification offers a simple mechanism to
> address several important modeling capabilities heavily used in almost
> all but trivial applications. Let me remind you of some of them using
> examples of some important tasks:
> 
> - aggregation: pull out a complete description of a book, incl. authors,
>      affiliation etc. from a dataset containing multiple book
> descriptions.
>      Need to know where to prune the graph.
> 
> - ordering: find the first author of a given book.
> 
> - nesting: attach a digital signature to a set of statements.
> 
> - quotation: enough of it on this list...
> 
> RDF folks seem in a tough stance trying to satisfy both XML folks and
> logic folks. XML folks want a data structure, and have very pragmatic
> and application-centric requirements. Logic folks want a logical
> language with well-defined semantics. While this balance is extremely
> hard to find, reification is one of the few mechanisms I'm aware of that
> looks like a data structure instrument (admittedly, a clumsy one), but
> is amenable to forging a logical interpretation around it (which might
> not be quite elegant either).

This is precisely the point that Pat and I have been arguing against.  We
haven't seen a logical interpretation for [uses of] RDF reification.  If
you can provide one, please do so.

> I have serious difficulties in explaining to developers why they should
> give up aggregation, ordering, and nesting that come "for free" with
> XML, and turn to RDF instead. I have really hard times finding
> satisfactory representations for these features in RDF, both from the
> perspective of programming convenience and model-theoretic
> interpretation. Maybe reification is a wrong hammer for those nails? If
> so, can anyone criticizing reification suggest a more suitable mechanism
> for handling the aforementioned features that makes both programmers and
> logicians happy?

The contention is that reification IS the wrong hammer for these nails.
There are logical formalisms that handle aggregation, ordering, and
nesting.  However, the semantics of the formalism has to incorporate these
notions, or provide some very high-powered semantic primitives---without
some semantic grounding there is nothing that can be done.

To consider just how hard some of these notions (that developers do indeed
want) are, just look at a full denotational treatment of the semantics of
programming languages.

> Sergey

Peter F. Patel-Schneider
Bell Labs Research

Received on Thursday, 17 May 2001 17:38:38 UTC