Schema mapping

At 09:06 PM 12/11/00 +0000, Bill de hÓra wrote:
[...]
>What would be nice, instead of embedding mappings ahead of time (you
>can't possibly guess all the mappings ahead of time), is given two
>schemas, a facility to create a third schema later on that posits
>mappings between the two. Sounds like a use case for out of document
>XLink processing. The number of such equivalence types may be small
>enough to get them into RDFS before it goes recommendation...
>
>Having a third schema to bind terms between two other schemas would
>would allow traversal all *across* hierarchies (this is also known as
>hyperlinking :). You could traverse across a series of such links to
>determine equivalent terms in-band. I guess you can call this out of
>document schema binding Metalinking.

I agree that a means to describe equivalences between schema would be 
useful, probably essential at some point (c.f. TimBLs comments on language 
mixing and "semantic schema" in 
<http://www.w3.org/DesignIssues/Evolution.html>).  I also believe it is too 
complex an issue to rush into...

I think equivalences may occur in several ways:
- direct resource/property equivalence;  renaming;  e.g. 'lessThan' <=> 'LT'
- universal equivalence of expression forms;  e.g. 'LT' <=> 'LE' && 'NE'
- equivalence under stated conditions;  e.g. 'A' => ( 'B' <=> 'C' )
- approximate equivalence;  good enough for some purposes  (1 Km approx 0.6mi)
- etc.

#g

------------
Graham Klyne
(GK@ACM.ORG)