[sc34wg3] Re: RDF/Topic Maps: what's an Application? (was: Re: [topicmaps-comment] RE: OASIS vs W3C)

[Steven R. Newcomb]

Tony.Coates@reuters.com writes:

> On 27/09/2001 23:37:30 "Steven R. Newcomb" wrote:
> 
> >I claim:
> >(1) A Topic Maps application can be an application of RDF.
> >(2) An RDF application can be an application of Topic Maps.
> >(3) Topic Maps itself can be an application of RDF.
> >(4) RDF itself can be an application of Topic Maps.

> Clearly there is a Turing Machine equivalence that
> applies here at the lowest level, so I can implement
> either topic maps or RDF using a suitably large
> abacus or a suitably organised barrel of monkeys
> (those little plastic ones I remember from my
> childhood).  

I take your point, but I'm wondering whether you missed
my point.  I was not trying to focus attention on the
possibility of implementing RDF using Topic Maps, nor
on the possibility of implementing Topic Maps using
RDF.  I was trying to focus attention on the questions:

* What would each of these very specific kinds of
  Applications look like?

* How would we declare each of these Applications
  formally?

* More generally, what provisions do/should RDF and
  Topic Maps make for the formal declaration of such
  Applications?

> A couple of more interesting questions,
> to me at least, are
> 1. How much more difficult would it be to implement a
> topic map application using an RDF engine rather than
> a topic map engine, and what would I gain/lose in the
> process?

> 2. How much more difficult would it be to implement
> an RDF application using a topic map engine rather
> than an RDF engine, and what would I gain/lose in the
> process?

These are great questions, but they can't be answered
in the absence of standards for the functions of RDF
engines and Topic Map engines.  (Or at least in the
absence of pre-selected RDF and Topic Map technologies
to be directly compared with each other.)

We must first establish conceptual foundations on which
such standards could conceivably be built, to which
implementations could actually adhere.  Then we'll be
in a position to establish the required functionalities
of RDF engines and of Topic Map engines.  Then, when
all that work has been done, we'll be able to answer
your very interesting questions.  We have to solve
these problems in their natural order, or we will waste
a great deal of time and effort, and, having wasted all
that effort, we will *still* have to solve these
problems in their natural order.

Personally, I believe that the very next question that
must be answered by the Topic Maps community is:

  "What constitutes an Application of the Topic Maps
  paradigm, and how should we declare such Applications
  formally?"

I believe that an Application of the Topic Maps
paradigm minimally consists of a set of association
templates (in other words, a set of assertion types).

* For example, if we were implementing RDF itself as an
  Application of the Topic Maps paradigm, we could
  declare an assertion type that exactly mimics an RDF
  assertion, with a "subject" role, an "object" role,
  and a "predicate" role.  RDF in a nutshell.

* Alternatively, if we were implementing a *specific
  RDF Application* as a Topic Maps Application, we
  would probably take a different approach, in which
  each of the types of RDF statements used in the RDF
  Application would become a distinct assertion type in
  the Topic Maps Application.  We would declare many
  assertion types, each corresponding to a specific
  type of RDF statement.

These two approaches are very different, and they offer
immensely different advantages and disadvantages,
depending on the nature of the information and the
usage scenario.  I think both approaches are correct
and viable.

The implications of each approach, for owners and users
of RDF information assets and of Topic Maps information
assets, are very different.  These implications should
inform us as we decide how to standardize an answer to
the question,

  "What constitutes an Application of the Topic Maps
  paradigm, and how should we declare such Applications
  formally?"

Ultimately, the answer to this question will have huge
impact on the answers to the questions you propose.

I would generalize your questions as follows:

  "What [Topic Maps | RDF (choose only one)]
  functionalities should be required of a conforming
  [Topic Maps | RDF] engine?"

  "How will conforming [Topic Maps | RDF] engines
  support *all* [Topic Maps | RDF] Applications?"

  "How will conforming [Topic Maps | RDF] engines
  support *specific* [Topic Maps | RDF] Applications?"

When the answers to these questions are known, we'll
certainly be able to answer your questions about which
kinds of engines should be used for which purposes, and
about the issues faced by information owners when
moving information between the two paradigms.

-- Steve

Steven R. Newcomb, Consultant
srn@coolheads.com

voice: +1 972 359 8160
fax:   +1 972 359 0270

1527 Northaven Drive
Allen, Texas 75002-1648 USA