RE: [xtm-wg] Issue with instanceOf

[Peter Jones <peterj@wrox.com>]

For reasons stated in other mails posted to the editors I do not believe the
type of 'scope' to be either topic or association or topic map as these
constructs are understood in TMs.
Nor do I believe it can be effectively modeled using combinations of these.
A scope is a different animal and requires a unique construct in the
processing model (whether that is there in the existing text or not I do not
yet know as I haven't read that bit yet).

So...bearing that in mind
I'll state my question in a different way.
Should the contents of instanceOf _ever_ be unscoped (whether unconstrained
or not), strictly speaking?

To dig back down to the fundamentals of the original little diagram that
Steve N drew in Paris...
When it comes to nailing the zero-dimensional point to the board we use
subjectIdentity.
And, IMHO, that is the only property of a topic that should be nailed and
that therefore should not be scoped; everything else should be allowed to
be.

Pete N wrote:
(2) is unnecessary because having a single association with multiple
types (each in a different scope) is equivalent to having multiple
associations (each in a different scope) where each association has a
single type.

Except that for some reason I might want to ensure that I have the same set
of endpoints (members) on an association to which I subsequently add many
types. Or a many-typed association that I want to add a new member too under
all the types of association at once.
Amongst other things the scope mechanism is a powerful compression factor.

Peter

-----Original Message-----
From: Daniel Rivers-Moore [mailto:daniel.rivers-moore@rivcom.com]
Sent: 04 January 2001 00:16
To: 'xtm-wg@egroups.com'
Subject: RE: [xtm-wg] Issue with instanceOf


Peter Newcomb's response points up the very issue I raised in my email
about syntax issues.

We should NOT be using the instanceOf subelement of association for the
purpose Peter Newcomb describes (which is to identify an association
that serves as a template for the current association). As Peter N
rightly points out, this is a different thing from instanceOf on topic.
Therefore (in accordance with the design decision already taken that
things that are different in nature should have distinct syntax), the
instanceOf element should not be used for this purpose. I have proposed
the use of an associationSpec subelement of association for this (by
analogy with roleSpec, which has exactly the same typing function with
respect to roles).

Peter Jones's question applies, however, to the ordinary use of
instanceOf, whether it be on on a topic element, an association element,
or an occurrence element. The purpose of instanceOf is to point to a
class of which a topic (or association, or occurrence) is an instance.

My position is that instanceOf is a syntactic shorthand for an unscoped
class-instance association. There may be merit in allowing this kind of
shorthand to be used for a scoped class-instance association as well.
This is the question I understand Peter J to be asking, and it merits an
answer. However, it's OK if the answer is to leave instanceOf unscoped,
because there does exist a longhand way of creating scoped
class-instance associations.

The longhand way of creating a class-instance association (scoped or
not) is to create an association element that has the public
class-instance association template as its template. However, given the
importance of typing to the topic maps paradigm, and the frequency with
which it is used, it is useful to have a syntactic shorthand such as is
provided by the instanceOf subelement.

Similarly, the occurrence is a syntactic shorthand for an association
between a topic and a reification of a resource, having the public
topic-occurrence association template as its template.

It is precisely these mappings between the syntax and the conceptual
model that need to be spelled out, and I have offered to create a first
draft of this. This task is intimately bound up with the declaration of
the relevant "public subjects". I was supposed to do this prior to the
Washington DC conference, but was unable to do so because there were
(still are) unresolved syntactic issues. I have raised these issues in
my earlier messages, and will allude to them again insofar as they are
relevant to the mapping between the syntax and the conceptual model,
which I shall be trying to produce in the next few dayus. 

Best regards

Daniel

-----Original Message-----
From: Peter Newcomb [mailto:peter@techno.com]
Sent: 03 January 2001 17:37
To: xtm-wg@egroups.com
Subject: Re: [xtm-wg] Issue with instanceOf


[Peter Jones <peterj@wrox.com> on Wed, 3 Jan 2001 14:34:27 -0000 ]
> why are the contents of instanceOf not scoped too?

In the case of an association element, the instanceOf element serves
to type the association.  If this typing relationship were to be
scoped, it would have to be done in one of two ways:

  1) by representing the typing relationship as another association in
     the topic map graph, or

  2) by allowing (in the conceptual model) scopes to apply to typing
     relationships even though they are not associations.

(1) is considered unacceptable because an association that models a
typing relationship between another association and its type must
itself be typed; if it is to be typed using the same mechanism as the
first association, it creates an infinite regression.

(2) is unnecessary because having a single association with multiple
types (each in a different scope) is equivalent to having multiple
associations (each in a different scope) where each association has a
single type.

-peter

--
Peter Newcomb                           Epremis Corporation
peter.newcomb@epremis.com               http://www.epremis.com/

To Post a message, send it to:   xtm-wg@eGroups.com

To Unsubscribe, send a blank message to: xtm-wg-unsubscribe@eGroups.com

To Post a message, send it to:   xtm-wg@eGroups.com

To Unsubscribe, send a blank message to: xtm-wg-unsubscribe@eGroups.com

To Post a message, send it to:   xtm-wg@eGroups.com

To Unsubscribe, send a blank message to: xtm-wg-unsubscribe@eGroups.com