Re: Picking the Tools -- Marrying processing models to data models

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

"Al B. Snell" wrote:

> Sub-type can be boiled down to "provides interface X" if you use proper
> interfaces / abstract classes. Providing the same interface is a pretty
> minimal requirement for "interchaneability". You don't have to inherit your
> proxy Customer from the original Customer object, you just have to provide
> the same interface.

Oh, my. Speak of the devil . . .This is precisely the argument that I just
referred to in my reply to Uche. My contention is that between the autonomous
nodes of the internetwork topology, implementing this sameness in interfaces is
undoable. Interface implementation requires the definition of a concrete data
structure, and that is precisely what autonomous nodes do not have in common
between their individual understandings of such abstractions as 'order',
'purchase price', or even 'scope of transaction'. The test of
interchangeability in that topology is whether the process defined at one node
can instantiate from the data structure presented to it something on which it
can operate, and then whether it can produce some useful outcome of that
operation. As a tool for facilitating such functionality, XML is a great leap
forward from OO, precisely because XML is text. An XML instance--the most
concrete realization for which XML itself provides--is still amorphous, even
abstract, with regard to the physical instantiation which it will be given by a
process. Put differently, the data structure exhibited by an XML instance is
still capable of sufficiently variable realization in process as to bridge the
lack of shared data definition between the autonomous nodes of the internetwork
topology.

Respectfully,

Walter Perry