Re: Abstract Type vs. Type Code

[Mike Rawlins <rawlins@metronet.com>]

Martin,

Thanks *very* much for the very thorough explanations.  I will base my review
comments on what I have now, but would appreciate continuing this discussion -
I think I understand what you're saying for each of the three cases, but could
you perhaps provide some examples?

Martin Bryan wrote:

> Mike Rawlins wrote
>
> :I don't think keeping a type code
> > in a superclass in order to keep a list of the subclass members is
> something we
> > want to do.  I'm sure that my experience in object modeling is limited
> compared
> > to several of your members, but I've never seen it.  Has anyone else?
> >
> > personally don't think keeping a type code
>
> Agreed, but first we need to realise that Type is being overloaded by the
> Core Components team.
>
> In the business process model there is something called Role that partners
> "play" in transactions. Core Components use Type to indicate role. If they
> changed the name of those places where Type is being used to identify the
> Role being played the link between core components and business processes
> would be much clearer.
>
> The second way in which Type is being used is that of "qualifier". This is
> intended to be used to distinguish specific subsets of a subtype which are
> not distinguished within the core component definition. All qualifiers are,
> in fact, names that should be applied to a set of context rules that apply
> to the subset for particular industries/process/etc.
>
> The third way in which type is being misused is to identify changes of
> datatype.
>
> In all three cases Type is used to identify a subsetting role that is
> affected by the context in which the parent element is being used. This is
> the key reason why your statement about dropping type codes is correct. It
> is a fundamental part of assembly (which assigns role names) and context
> rules (which idnentify qualification and datatype restrictions.
>
> Martin Bryan

--
Michael C. Rawlins, Rawlins EC Consulting
http://www.metronet.com/~rawlins/