[
Lists Home |
Date Index |
Thread Index
]
> Is everyone who wants datatypes comfortable with a
> minimum set or is the perennial "how to extend this
> in a standard way" argument being had?
No. I want to be able to define any simple type I want. This includes:
1) types based on mathematical axioms and proofs
2) types whose values are determined by laws of nature (empirical and
repeatable)
3) types based on rigidly defined international standards
I also want:
4) types based on specific locale
All of these would have clearly defined membership (value spaces).
Some of these values spaces would not be computable.
Of those that are computable, some would be prohibitively expensive to
validate.
All, however, would allow the unambiguous communication of a concept.
All types must have at least one lexical representation to be meaningful.
A type may have a universal lexical representation, or it may not.
A universal lexical representation must be capable of depicting all values
of the type.
A type may have both universal and local representations, or only local
representations.
Local lexical representations may only allow depiction of _some_ values of
a type.
For those types with universal representations, a local representation _may_
have a defined mapping to the universal representation.
For types lacking an accepted universal lexical rep, then each local rep
_may_ describe mappings to other local reps.
A local rep _may_ have only a partial mapping to another local rep.
Existence of a universal lexical space does not prohibit direct
locale-to-locale mappings.
All respresentations of values in a locale must have a provable mapping to
one and only one universal rep (if it exists). Implementations of such
mappings do not have to be provided or made available.
Local representations can be intersected with where partial mappings are
available. The resulting lexical and value space must be well defined.
--------
In other words, types first and formost define a concept of membership.
Such definitions must be formal and unambiguous.
If it's validatable, great.
If it has a universal representation, great.
If it has one or neither, it's still useful for articulating what a data
value _is_.
|
