A not on types (was Re: [RIF] homework for 10/17 telecon)

On Oct 16, 2006, at 7:18 AM, Michael Kifer wrote:

>
>
>> Michael Kifer wrote:
>>
>>>> In any case we need IRIs for the relation and function symbols
>>>> irrespective of sorting.
>>>
>>> No, this is the first step in adding sorts.
>>
>> No, surely it's the first step in webizing[*] a language.
>>
>> Dave
>
> You can put it this way, but IRIs and other data types are nicely
> formalized as sorts. So, this is the most natural way to approach  
> these
> issues (incl. webizing). I thought it was clear, but if not I hope  
> that
> this discussion clears things up.
>
>
> 	--michael
>

As I have learned from trying to 'do' types for the last 10+years,  
handling types properly in non-trivial. (Waving the sorted-logic wand  
over a set of rules does not cut the mustard.)

In any case, the RIF is supposed to be an *interchange* format, not  
yet another rule language. That means that it is not enough to come  
up with a single type system (a point solution in the multi- 
dimensional type space) but a way of capturing types and type systems  
that are already in use.

Frank

Received on Monday, 16 October 2006 17:56:34 UTC