Re: Object Role Modelling (ORM) or UML or ?? for designing Schemas

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

Ken North wrote:

> > "...  My guess is that the syntax-centric
> > stranglehold will not be broken until there's a conscious focus
> > on conceptual modeling, accompanied by an unspoken agreement
> > that XML schemas and other markup syntaxes can readily be
> > generated from conceptual model notations..."
> >
> > but the comment appeared to have fallen on deaf ears.
>
> My sense is quite a few developers are on the same wavelength.

[snip]

> There is clearly an advantage to using the same model to generate XML and
> database schemas. However, people sometimes have to work in a manual
> syntax-editing mode before they appreciate the productivity of tools.

And then there are the few of us who are committed to markup precisely because
syntax is its inescapable, irreducible nature. I believe that we should now be
designing processes which are inherently well-suited to operate in an
internetwork topology. An internetwork is radically different from the
homogenous 'enterprise' network topology which is fundamental to the SQL world.
The two salient characteristics of the homogenous network are that it is closed
and that the functionality of each of its nodes is transparent to the others. A
network is closed because in order to participate a node must conform in its
processes to the shared, a priori data schemas which unify that network or, if a
node is outside that network by virtue of adhering to the data schematics of a
different organization, its every transaction must be transformed at a gateway
to bear the semantics expected within the network with which it communicates.
The advantage of this conformity is that every node of the network has a
comprehensive understanding of the functionality of the other nodes, based on
the semantics implicit in their enforced uniformity of data schema.

An internetwork on the other hand is constructed not by conforming the data
understandings of its constituent networks to some single agreed standard, but
by simply overlaying a common addressing scheme across the participating
networks. Within those networks, and even at individual nodes, the understanding
of data--as expressed first by the structure in which it is cast and second by
the manner in which it is processed--is entirely local and autonomous. The
internetwork provides the advantage that nodes once out of reach because they
were outside the homogenous network may now be contacted, and thereby act as
counterparties to transactions, by virtue of the common addressing scheme. They
cannot, however, be expected to share the fundamental, but local, understanding
of data on which the form and nature of transactions on the homogenous network
is predicated.

Communication between such autonomous networks and nodes should be inherently
well-suited to the nature of the internetwork topology which makes that
communication possible. It should not, therefore, depend upon the purely local
semantics which are expressed by a particular schematic or understanding of
data, nor upon a particular form of processing which follows from that same
understanding of such data. As syntax, markup offers that basis of communication
when, and if, it is clearly understood that the semantics elaborated from that
syntax will be entirely local, and different, at each autonomous node which
might perform some useful process against that data.

Respectfully,

Walter Perry