[sc34wg3] SAM 3.4.4 Reification and 3.4.5 Properties

Lars Marius Garshol sc34wg3@isotopicmaps.org
15 Apr 2003 21:35:59 +0200

* Luis J. Martinez
| Are the properties [subject addresses] and [reified] representing
| the same thing when the subject is an information item in the topic
| map?

Good question. The answer is 'no'.
| The component of Subject seems to be implied by the properties of
| [subject identifiers], [subject addresses], and [reified]. Subjects
| are an very important part of Topic Maps. Why don't we have a
| separate component called Subject that can be compose of information
| items, subject indicators, or AbstractTopicObjects. Is it possible
| to model the Subject "component" instead of implying it?

If you look carefully at the SAM you'll see that the prose divides it
into three layers:

  a) the subject world,

  b) the topic world, and

  c) the SAM representation of the topic world.

Basically, the subject world is the world in which you and I live.
The topic world is the world of topic maps (abstractly conceived), and
the last layer is that of the SAM representation of topic maps (still
abstract, but no longer quite as abstract). Every layer represents, or
models, the layer above it.

In topic maps, subjects are represented by topics and topics in turn
are represented by topic items. So subjects in one sense are already
in the SAM, through the topics that represent them.

We could of course add subject items to the model, but they would then
differ completely from all the other item types in the model. They
would be members of the subject world intruding into a lower level
that already has a representation of them.

That's not to say that your suggestion is wholly wrong. I think for a
conceptual model that would make sense, but the SAM is a data model
designed to support interoperability (XTM/HyTM/CXTM) and the
definition of query and constraint languages.
| Can we describe the AbstractTopicObject component separately with
| its properties including [reifier] and express that is a
| generalization of TopicName, Variant, Occurrence, Association, and
| AssociationRole? That would make the model cleaner and clearly show
| that all of the above mention items can be reified by a topic
| instead of repeating the [reifier] property.

We could do that. In fact, the authors have discussed this several
times and have found that this is a trade-off between brevity of text
(good) and simplicity of machinery (also good). We would need to
introduce inheritance in the information set metamodel, which does not
have it, and have been reluctant to do so.

If more people had made the same suggestion as you we would probably
have adopted it, but you are actually the first person to suggest
this, which implies to us that settling for simplicity of machinery
(no inheritance) was the right choice.

If you have good arguments for why it is not right we are of course
willing to hear them. :)

Lars Marius Garshol, Ontopian         <URL: http://www.ontopia.net >
GSM: +47 98 21 55 50                  <URL: http://www.garshol.priv.no >