[sc34wg3] tmrm-map-inferred-information

[Lars Marius Garshol]

* Patrick Durusau
>
> I assume that the only inferencing that we are concerned with is  
> that of supertype-subtype?

Not necessarily. There is subtyping of topics, subtyping of  
statements, and then there is the issue of what is implied by scope,  
etc. I think we agreed to limit ourselves to the two kinds of  
subtyping in Leipzig.

A full formal semantics would require more than that, though.

> Since it is transitive it is possible to have relationships that  
> are not explicitly represented in the topic map.

Correct. And the same thing is in principle what happens in all the  
other cases, too.

> Which means that a processor the supports inferencing will have to  
> deal with loops. (TMDM 7.3, Note 1)
>
> And, "...should be interpreted that the sets of all instances for  
> all types in the loop are the same."

Yes.

> I am having trouble with that part of Note 1. Does it mean that all  
> the instances are instances of all the types, even though the note  
> goes on to say that does not imply that the types are the same?

For all t where t is a type in the loop, the query

   $i isa t

will produce exactly the same result. However, the t's will actually  
be different topics.

Does that clarify it?

> In other words, if there is a loop, I could infer that some  
> instance is an instance of multiple types?

Yes.

> I don't have any suggestions but just wanted to make sure I was  
> understanding the issue properly.

Well, this is not the issue. This is not an issue at all, actually. :-)

The real issue is, let's say we have the following topic map (in CTM):

   a isa b .
   b ako c .

If I now do a query to count all the associations that 'a' has, is  
the answer going to be 1 (a isa b) or 2 (a isa b, a isa c)?

I've added this explanation to the issue now, in the hope that this  
makes things clearer.

--Lars M.