Re: [topicmaps-comment] occurences and modeling resources

[Lars Marius Garshol <larsga@garshol.priv.no>]

* Howard Goldberg
|
| 1.  In the conceptual model, an occurence is uniquely identified by
| an instanceOf, a scope, and a resourceRef.  Therefore, it appears
| that any occurrence that shares the same tuple should have the same
| occurence identifier.  

If two occurrences of the same topic have the same values for these
properties (as well as for the <resourceData>, if they have any) they
need to be replaced by a single occurrence. (XTM 1.0, annex F.2.6.) 
I am not sure if this is what you are thinking of, however.

This is not the same as saying that the identity of an occurrence
object in the internals of a topic map engine is necessarily
determined by these four values. There is no rigorous specification of
what is required here, nor do I think that there should be.

What you mean by "occurrence identifier" I don't know. Could you
explain? Your question seems to be about this identifier, so it sounds
important. 

| For example, if I have multiple topics that each appear in the
| english version of my home page, each occurence in every topic
| shares the same identifier, and are reified by one topic whose
| subjectIndicatorRef is that single occurence identifier.

Here it seems that you are making a common mistake: the resource
referred to by this occurrence is *not* the same as the occurrence.
The resource is just the resource, while the occurrence represents the
relationship between the resource and the topic.

So it seems that you want the following:

 - a topic for each subject discussed on your home page (actually, you
   probably don't want this, as the number of subjects is likely to be
   enormous) 

 - an occurrence of each of those topics referring to your home page

 - a topic that reifies the home page, with a <subjectIndicatorRef>
   pointing to the home page

Does this sound right?
 
| 2.  If my home page appears in several languages, the reification of
| my home page as a resource must be a tree.  

Not necessarily. You could decide that there is a single abstract work
(my home page), of which each of the language versions is a concrete
representation in text, and thus just make a single topic for the home
page. Or you could do both, or you could just do one topic for each
version. What is best depends on what you want to do.

| The reification of each language version of my home page becomes a
| single topic whose subjectIndicatorRef identifies the occurrence.

If you replace the word "occurrence" in that sentence with "resource"
you've got it.

| However, each of these language-specific occurrences is an instance
| of a topic reifying my home page, the subjectIndicatorRef of which
| could be some unique URI.

(Again, "occurrences" -> "resources".) There is obviously a
relationship between the language versions and the abstract work, but
that is not the class-instance relationship. I'd say that the
relationship is better described as "concretization of", or perhaps
"representation of".
 
| Intuitively, I'd like to model a resource and its multiple scopes as
| a single topic, 

Well, resources don't have scopes. In a topic map, what has scope is
only assertions about subjects, and a resource is not a topic map
assertion about a subject, it's just a resource. Occurrences have
scope because they assert that the resource has some relevance to the
topic, and that relevance is qualified by the scope.

| This seems to put the burden on the topic map processor, in order to
| support interesting queries about resources, and the tool writer, to
| hide this complexity from users who want to create resources on
| which to map occurrences of other topics.

Here you lost me. Could you expand on this, or perhaps give an
example?

--Lars M.