Prolog and XML (Re: NPR, Godel, Semantic Web)

[Rick Jelliffe <ricko@allette.com.au>]

On the subject of Prolog, it seems that prolog systems have a distinction
between compiled rules and inferences that may not fit well into dynamic
systems.

One of the problems with many SGML systems was that they required a
compile phase (rules building) separate from using documents: this was
clumsy and made the system feel complicated--also it created different class
of users: the gurus versus the drones.

But XML documents with arbitrary namespaces seem to prefer incremental
loading of rules.  At least one prolog system I have seen warns that, while
it
can be done, it is expensive.

My questions are these:
 1) if incremental loading is expensive in Prolog systems, how expensive is
it really? is there some heuristic that can guide us to figuring out the
maximum size of ruleset in which incremental loading would be tolerable?
 2) are there Prolog (or Prolog-like) systems available  which allow
efficient
incremental loading of rules?  (I'd guess that a Java-based system might be
more
likely to provide this than a C++ one with a native compiler.)

Cheers
Rick Jelliffe