Re: Core Component Analysis - SWIFT's Comments

[Martin Bryan <mtbryan@sgml.u-net.com>]

Jacques, Nigel et al

First let me start by thanking Jacques for his excellent paper. It has one
or two minor errors, the most critical of which is the claim that
dependencies are not allowed for in the metamodel (see 1.3). They are, but
they are not recorded in the entity use area but  in the Context related
parts of the model. (They are there because they are modifiers of the
datatyping rules and models.)

Now let me come on to a point Nigel made:

> Our core component dictionary can then concentrate on defining and
> agreeing semantic meaning for common business data e.g. Transaction
> Effective Date - it is important that we have an agreed understanding of
the
> meaning for this data.

It is important that we realise that there are two types of "core
components": abstract core components and context-specific core components.

There needs to be an AbstractDateTime component that forms the basis for all
dates. This should form the basis for specific applications of dates, such
as the TransactionEffectiveDate required by Finance and Insusrance
applications. In other words, TransactionEffectiveDate is a synonym for
AbstractDateTime that applies to a named st of "domains". The same goes for
AbstractParty, which has synonyms such as Buyer, Seller, Client and Insurer.

It is important that we clearly identify when a context-specific core
component is really a synonym for an abstract core component. It is also
important that we record the constraints that apply to the use of an
abstract core component within a particular context. For example, there may
be a constraint that times may not be recorded as part fo the
TransactionEffectiveDate, even though they form part of the definition of
the AbstractDateTime. It was for this reason that the complete data capture
forms allowed for the assigning of both "local names" and constraints when
applying an abstract core component within another core component.
Unfortunately the switch to spreadsheets has "lost" this functionality.

We still need to develop a language for describing model constraints.
Unfortunately this work is running behind schedule. But the analysis team
must work on the presumtion that a suitable technique will be defined by the
end of the Vancouver meeting.

Martin