Christian de Sainte Marie wrote:
>
> Or replace them all by an abstract construct that could be specialised
> into Groups, Rules, Facts, Whachamacallit etc in concrete dialects;
To make my point more clearly, and, hopefully to help progress the
discussion, here is a diagram 5 of FLD with only abstract classes for
the content of a Document: a RIF Document contains zero to many objects
of the abstract class MACHIN, which can be subdivided into objects of
the abstract class TRUC and objects of the abstract class CHOUETTE.
A Document can have metadata attached, as does any (concrete) object of
a MACHIN class as well, but this might actually be orthogonal to the
discussion (as I suspect Paul is arguing in the part of his email that I
did not read yet :-)
The main difference between the TRUC and CHOUETTE (abstract) subclasses
of MACHIN is whether they recurse (on MACHIN) or not: TRUC does not and
CHOUETTE does. So, TRUC concrete classes are FORMULA classifiers,
whereas CHOUETTE concrete classes are classifiers of FORMULA grouping
concepts.
Another difference is that TRUC (concrete) objects are directly
associated to a FORMULA, whereas CHOUETTE objects are associated to
groups of FORMULAs only via the concrete classifiers of the TRUC class,
if any.
Now, what if FLD could specify only these abstract constructs, with
whatever names seem most appropriate, and leave it to concrete dialect
to specify the concrete classes they need?
One dialect, say: BLD, could use one CHOUETTE concrete class called
Group, and two concrete TRUC classes called Rule and Fact; whereas
another dialect, say: PRD, could have only one concrete TRUC class
called Rule and no concrete class for CHOUETTE at all.
Yet another concrete dialect could have only one concrete CHOUETTE class
(Group) and no concrete TRUC class: Group vould thus contain zero or
many Groups or FORMULAs, as is the case in the current draft of BLD and
FLD...
Would something like that work? Or be amenable to work?
And would it help resolve the ISSUE 51?
Cheers,
Christian
