Index Cards RE: [sc34wg3] Modularization

[Martin Bryan]

Bernard tried to use the library index metaphor for topic maps:

> In the simple cosy world of ancient (and not so ancient) libraries, you
> had books and index cards. An index card for each book, with title,
> author, DDC classification, format, ISBN number, date of purchase,
> current status (available, borrowed by reader X) and so on.

The problem with this proposed model is that libraries do not use a single
index card: each book has multiple cards, one filed in the author index, one
in the subject index and perhaps one in the titles index. Also libraries can
have multiple copies of each referenced book. Note that there can be many
authors with the same name (e.g. John Smith), many books with the same name
(e.g. The Holy Bible) and many subjects with the same name (e.g. carriage)
which is used in different contexts.

So

> 1. SLUO : Having exact one-to-one correspondence between one index card
> and one book.

is not ot valid. The SLUO applies to the subject, not to the object. Where
there are multiple occurrences then there is not a one-to-one
correspondence.

> Or, if you have several cards, have a way to merge them.
> Every good librarian knows how to do that.

Again, not valid. A good librarian has to distinguish things that should not
be merged by applying scope to information. Cards depend on context for
their use. A classic problem is whether or not all the titles of a
particular author should be listed sequentially when there is more than one
author of a name, or whether all the titles prepared by authors with a
particular name should be listed alpahbetically without any way of
identfiying cross relationships.

> 2. Assertion of relationships between cards (staples)
> How do you express in a standard way relationships (associations)
> between subjects, using formal binding of matching cards?

A librarian defines relationships by having standard fields (roles/type) to
classify data (i.e. using a standardized classification scheme to define the
relationships between groups of related entries)

> 3. Structure, format, content of the cards (what information needs to be
> there, under what form, what colour cards need to be, and how you manage
> card boxes)

We must not forget the three things that distinguish topic maps from other
data classification schemes:

 - scope
 - role
 - the ability to define sets of occurrences (and names) for a single topic,
rather than one-to-one relationships

These are our strengths. Without them RDF would suffice.

Martin Bryan