[sc34wg3] revised draft Reference Model document N0298

[Steven R. Newcomb]

"Graham Moore" <gdm@empolis.co.uk> writes:

> If I have a topic with a subject indicator 'urn:gdm'
> then in the reference model that gets turned into two
> topics the topic concept of 'graham moore whatever
> that is' and the concept string 'gdm:urn'.

You haven't explained your example completely, so I
don't know exactly what your scenario is.  There may be
either two or three topics here, and I can't tell which
situation applies.  It all depends on the answers to
the following questions:

> If I have a topic with a subject indicator 'urn:gdm'
              ^^^^^

Question #1: What is the subject of your topic?  Is the
             subject a name, or is it Graham Moore, the
             living human being?

> If I have a topic with a subject indicator 'urn:gdm'
                                              ^^^^^^^

Question #2: A subject indicator is always a piece of
             addressable information -- a subject
             constituter.  It is inherently a subject
             constituter, but it is a subject indicator
             only by virtue of the fact that a topic
             map author asserts that it is one.  When a
             topic map author asserts that that piece
             of addressable information is to be
             interpreted as a subject indicator,
             presumably the author regards it as a
             compelling, precise, unambiguous indicator
             of some specific subject.  Usually, it's
             necessary for a topic map user to
             understand the *context* of a subject
             indicator in order to understand what the
             subject is that is supposed to be
             indicated by it.  In your example, your
             subject indicator is apparently a string
             without any context at all.  Therefore, I
             can't tell how I'm supposed to understand
             or interpret it.  Is it a name?  A part
             number?  A word in some natural language
             that I don't happen to know?  I have no
             way of knowing, because you haven't
             specified the context of this addressable
             data.  (If the context is your e-mail, to
             which I am now responding, then I must
             tell you that your choice of subject
             indicator for your topic is ambiguous!)

Depending on your answers to the above questions,
there may be two subjects:

  (1) Graham Moore the human being, and

  (2) the addressable subject which is the string
      'urn:gdm' in some particular location (which your
      example hasn't specified).

  ... where (2) is asserted to be a subject indicator
      of (1).

Or, there may be three subjects:

  (1) Graham Moore the human being, 

  (2) one of Graham Moore's names, which is "urn:gdm", and

  (3) the addressable subject which is the string
      'urn:gdm' in some particular location (which your
      example hasn't specified).

  ... where (3) is asserted to be a subject indicator
  of (2), and (2) is asserted to be a name of (1).

> They are connected by one assertion of type 'subject
> indicator'.

Yes, but which two topics are so connected?

> Question is what happens when another topic of a
> string value 'gdm:urn' is introduced.

What is meant by "string value"?  Names and subject
indicators are not the same things.  Names are *always*
nonaddressable subjects (and, like all nonaddressable
subjects, they can have subject indicators).  Subject
indicators are *always* addressable subjects.

> According to the reference model the two 'string
> value' topics are merged.

Whether they are merged or not depends on how you
answer the questions at the top of this note.

> What I am saying is the the two topics
> identified by the subject indicator should be merged
> - not the string topics.

What is a "string topic"?  Is its subject a name, or is
its subject a piece of addressable information?  It
can't be both.

> Maybe this isnt the case and identity is in fact a
> property of the topic - but from the diagrams and
> prose thats not how it appears.

> Thus subject identity properties should not be
> seperated from the topic. I idenity as being an
> inherent property of a topic.

The draft Reference Model proposes only one inherent
property for topics.  It's a property that only those
topics that have addressable subjects can have.  For
each such topic, the value of this property is the
addressable subject itself -- the addressable piece of
information, considered in its own context.  According
to the draft Reference Model, all other things that we
might consider to be "properties" are actually "roles
played in assertions".

> I've attached the GIF again steve.

Many thanks.  I see it now.  (I'm still getting used to
the mind-boggling power and complexity of the Gnus
mailer, which I think is probably the reason I didn't
see your attachment before -- "operator error".)

As I look at your GIF, my suspicion that you're
confounding names with subject indicators is
reinforced.  There are more different subjects than
you're recognizing, I think.  On the other hand, I'm
really not very certain that I'm interpreting your GIF
the way you intend.

-- Steve

Steven R. Newcomb, Consultant
srn@coolheads.com

Coolheads Consulting
http://www.coolheads.com

voice: +1 972 359 8160
fax:   +1 972 359 0270

1527 Northaven Drive
Allen, Texas 75002-1648 USA