From: Ray Gates (home) [gatesray@home.com]
Sent: Monday, 2001 June 11 9:46 PM
To: Dan Gillman
Cc: WG2 News
Subject: Initial Comments on PDTR 20943-3

Dan,

I have started to review PDTR 20943-3 Content Consistency for Value Domains,
and thought I would provide some initial feedback in time for this week's L8
TGTD meeting.

In general I like what I see so far, but there are a few points I disagree
with, and which I believe deserve some discussion.

Topic 1: Equivalence of Conceptual Domains
------------------------------------------
In section 4.1 you have given good examples of different value domains of
codes representing human sexes.  However, following Example 3, you state:
"All the value domains for human sex codes have the same conceptual domain."

I would argue that for enumerated domains, a conceptual domain consists of a
set of value meanings, and that if two sets of value meanings are not
identical, they must by definition represent different conceptual domains.
Your examples 1 and 2, each represent a set of three value meanings, so the
two value domains are representations of the same conceptual domain.  Your
example three has a set of five value meanings, and therefore is necessarily
a different conceptual domain, albeit a superset of the other. We might name
the three-meaning domain "Categorization 1 of human sexes" and the
five-meaning domain "Categorization 2 of human sexes". [Someone who actually
uses multiple such categorizations might have more meaningful names.] The
metamodel supports a relationship between conceptual domains where the
subset/superset relationship could be described.  If this comment is
accepted, then item 6) of page 6 also needs to be changed to delete the
reference to Human Sex Codes (3).

Topic 2: Value Domains and Classification Schemes
-------------------------------------------------
The suggestion that Value Domains can represent Classification Schemes is an
interesting one, but something that is not explicitly supported in the
model.  To support it explicitly, I think there would need to be a
relationship between Classification Scheme Item and Value Domain.  [One
could also argue that at another level, csi-value is a data element that can
have an associated value domain.]

Topic 3: Distinguishing Data Element Concepts from Conceptual Domains
---------------------------------------------------------------------
In item 8) on page 7, you have presented a description of the distinction
between data element concepts and conceptual domains.  While this appears to
be correct, I would like to try an alternative approach which relates back
to the examples.

"A 'data element concept' represents the concept of a property of a specific
class of objects (e.g. sex of employee, or sex of student) while a
conceptual domain represents the concept of the property independent of
those particular objects (e.g. human sex)".

Of course, if I define an object class "Human Being", with a property "Sex
of Human Being", it becomes hard to distinguish the data element concept
from the conceptual domain. Is this a weakness in the above description
only, or does this point to a lack of clear distinction between the
concepts?

Topic 4: Versions of Value Domains
----------------------------------
In item 2) on page 5, you state that each time new permissible values are
added or subtracted, a new value domain is created.  I believe (and hope)
that a registrar actually has the choice of creating a completely new value
domain or a new version of an existing domain.

That's all for now.  More later.
Regards,
Ray Gates