Re: RM formalism supporting partial information

On Nov 16, 10:30 am, Jan Hidders <hidd..._at_gmail.com> wrote:

> To give a hint about my opinion on the final question. If you assume
> the closed world assumption, and we usually do in this context, then
> it is not true that a smaller relation necessarily contains less
> information. The absence of tuples then also carries information.
>
> So far for now.
>
> -- Jan Hidders

Here are two good examples from Atsushi Ohori and his paper "Ordering and types in databases". Hope they can help to clarify things. First example is about Zaniolo's ordering. A. Ohori explain it as ordering of database objects regarding description. DB objects describe some real world objects. " These incomplete descriptions are partially ordered by how well they describe real-word objects."

The first example is something like:

     ( Name -> 'John Smith' , EmpNo -> 33 , Sex -> 'M' )

<= ( Name -> 'John Smith' , EmpNo -> 33 , Sex -> 'M' , Age -> Null )

<= ( Name -> 'John Smith' , EmpNo -> 33 , Sex -> 'M' , Age -> 55 )

The second example is like

(Name : string , EmpNo : Int , Sex : string , Age : int )

<=1 (Name : string , EmpdNo : int)

Here the smaller relation contains more information. A. Ohori explains this as " Since more general means less informative..."
However ordering (<=1)in example2 is not Zaniolo's ordering, this is ordering on types.
For more details and notation see above mentioned paper. Now regarding example2 you can try to define partial order on interpretations.

Vladimir Odrljin Received on Mon Nov 19 2007 - 17:56:27 CET