[topicmapmail] Processing Model issues

[Piotr Kaminski]

Hi,

> No, it is not (a good idea). The problem is that you can't escape this
> infinite "semantic recursivity" if you don't define clearly a
"meta-level".

I agree.  I think the trick to having a flexible system, though, is to make
this primitive meta-level as thin as possible, and to provide reflection
facilities that allow you to express it in "normal" terms.  I'm not
convinced the topic map model has picked the best primitives.

> distinct from the "instanceOf". That is what the seeked template mechanism
> should be IMO. It's the only way I see to escape infinite
> recursivity in a "non-lazy fashion".

The template mechanism (at least the proposals I've seen) won't help escape
the recursion.  The templates simply add more information to the association
types, but the instanceOf relationships are still required.

A way to escape this infinite regress would be to change how the type of an
association is indicated.  Instead of using a typed association, let the
type topic be an association itself that contains all its instances as
members (playing the "instance" role, for example).  Since members/roles are
primitive concepts that don't have types, we avoid the regression.

Of course, a good implementation of such a system would need to provide
reflective facilities so that the members of an association could be viewed
as independent topics, associated to the association topic with associations
that had "association", "member" and "role" roles.

        -- P.

--
  Piotr Kaminski <pkaminsk@who.net>  http://www.csr.uvic.ca/~pkaminsk
  To do:  put witty quote here.