>As an aside, I note that much of this discussion revolves around issues of
the definitions of key terms in the discussion, which is, of course, one of
the purposes of topic navigation maps: to define things.
I have long stressed the need for definitions to be primary objects in Topic
Maps, but have consistently been battered down on the grounds that you don't
need definitions for every topic (something I am far from convinced about,
especially for maps that are to have a life of more than a few months).
The idea of public resources that contain definitions is an acceptable
alternative to definitions helped me accept the lack of priviledged
definitions. It means that topic maps can share definitions by simply
referencing the same set of definitions, so that you could automate
relationships between maps. For example, all maps that reference Library of
Congress concept XYZ can be considered to be indirectly related to each
other. The interesting point is that these same thing can be said about
things in Dewey classification 888.999.000 if this can be defined in the
same manner. The question is whether a single way of referencing the object
is adequate, or whether we should accept that definitions may be known by
multiple public identifiers.
To me it is paramount that we provide a mechanism for referencing the
definitions of topics in a way that makes topic maps interlinkable, and in a
way that allows multilingual data sets to be maintained over long periods of
time. If we can do this by use of sharable ontologies then we have a chance
of linking object sets that otherwise must irrevocably be kept asunder due
to the fact they use different names to describe the same concept.
Martin Bryan