Being flexible about categorization

These quotes from Kelly Green caught my eye from an actKM discussion:

Information systems require clearly delineated properties. I think many people involved in this space would agree that this is an artificial boundary and that the boundaries in "real world" are more elastic.
...
Ontologies are good for illuminating subjects -- but they seem to also constrain them.

Maybe the Web world needs a space for "fuzzy categories" -- call them "associations" or "resemblances"? Where a topic can be 70% related to 'fluffy bunnies'?

I know it's not a new idea, but particularly with the frenzy of tagging lately, perhaps we need a more nuanced approach to categorization.