Well-Behaved Ontologies (was: n3 wishlist: automatic namespace selection)

> > From: Sandro Hawke [mailto:sandro@w3.org] 
> > I was thinking there would be no ordering -- if a definition 
> > was found in more than one ontology, an error would be 
> > reported.  If ontologies were well-behaved in never losing 
> > terms, this means at worst one's code
> > would produce an error instead of the wrong results.    Is that still
> > too dangerous to provide to users?
> 
> Another problem is ontologies *gaining* terms - you might suddenly find that
> a previously well-behaved system fails simply because someone has introduced
> a term to an ontology that has an identical name to one you're already using
> (with its default name) in another ontology.  This *will* happen on the
> Semantic Web, because people won't be as careful as we would wish about
> versioning and never replacing the contents of a particular URL.

I wasn't very clear.  Yes, I was assuming "well-behaved" ontologies
would only gain terms, not lose any.  And I went on to say that this
kind of change would produce a different (much less dangerous) kind of
error, and therefor might be acceptable.

But now that I think about it, I'm really not sure what changes to an
ontology are legitimate.   There are really two kinds of changes:
adding constraints and removing them.   Is one of these somehow better
for users?

There's a lot to be said for forbidding changes at all, importing a
particular version.   

    -- sandro

Received on Friday, 8 March 2002 10:35:26 UTC