xml-dev - Re: Property role and domain

["W. E. Perry" <wperry@fiduciary.com>]

My own view is that this multi-dimensional modelling of data is, par excellence,
the task of XLinks. Neither role nor domain is static, nor are the relationships
in the data instance among simultaneously applicable roles and domains. The
simple ontology presented by element markup must itself be reinstantiated afresh
at the document-consuming node with each new use of the data instance. The
specifics of that instantiation will necessarily depend upon the local
environment of the node performing the instance processing:  that
document-consuming node is simply not the same as the document creator, and the
form in which, for its own unique purposes, it is required to, or capable of,
instantiating a received document is not only locally unique but unknowable in
detail to the document creator. Far more so than the ontology expressed in
simple element markup, roles and domains are intimately contingent upon the
local context. Their very nature lies in perspectives on the relationships on
data marked up within the instance document, and those perspectives require an
observer, and a process, which is inherent and unique to the local node.

The real issue here is the expression of likeness, sameness, or even identity,
when the local form which that sameness takes will differ from node to node in
ways unique to each local context. To use a part of your own example, the nature
of 'billing' or 'shipping', most particularly when expressing a role or a
domain, is utterly different depending on whether one is the originator or the
recipient. Precisely what differs is the process which the local node must
perform, depending on its role, and the definition of that role in markup should
properly be done by reference to the (inherently local) process itself. That is,
the sender and the recipient will both call the process 'billing' and in doing
so will express the fundamental sameness of the operation. But the process which
each node performs will be utterly different and ultimately dependent upon the
local context. At each node, XLinks using the same or equivalent terms should
refer to the local version of the process to be performed. They will thereby
define the role or the domain in the only practicable way in which it can be
defined within a distributed topology of autonomous nodes--by reference to the
process which determines the unique local instance instantiation of that role or
domain.

Respectfully,

Walter Perry


Kay Michael wrote:

> This need to represent both the role of a property and its domain has always
> been a problem in data modelling, for example the concept of domain was in
> Codd's original relational model but didn't make it into SQL until much
> later, and never made it into the core of the language.

[snip]

> Any other views on this? Does it become easier with XML Schema replacing
> DTDs?


***************************************************************************
This is xml-dev, the mailing list for XML developers.
To unsubscribe, mailto:majordomo@xml.org&BODY=unsubscribe%20xml-dev
List archives are available at http://xml.org/archives/xml-dev/
***************************************************************************