ISSUE-5: n-ary datatypes - use cases - preferred design

Jeremy Carroll wrote:
> also on what I think the high-level user goal is, 

Use cases that occur in the literature for n-ary datatypes is adding 
simple arithmetical facts to an ontology like:

- unit conversion
- simple limits on volume or permiter

This is done by means of a 'datatype' which works with two or more 
datavalued properties and is true for an individual if those properties 
have values for that individual, and the values satisfy some constraint 
represented by the datatype.


===

My preferred design is that such aspects are processed in a procedural 
way, for example, during data loading, or perhaps using some sort of 
procedural attachments to a knowledge base.

For example, unit conversion can be done by deciding up front to work 
with the metric system, and systematically performing data laundry at 
load time, on values using the US system.

While data laundry is known to be a pain, it also has known strengths 
and weaknesses.

The same approach to a simple limit, would be to, on data loading, 
compute some qualitative measure "small box" "medium box" "large box".

While both of these examples are much weaker than with the treatment 
using the n-ary datatypes design, I believe that this is sufficient for 
currently known needs, and does not require more research.

Jeremy

Received on Monday, 12 November 2007 16:43:01 UTC