Taxonomies are a way of pigeonholing information - a place for everything, and everything in its place. Classic examples are the Linnaean taxonomy of living things and the Dewey Decimal Classification system used in some libraries.
Matthew Hodgson has a good article on Taxonomy as a form of information classification that puts an Information Architecture slant on taxonomies.
When I started doing Information Architecture work I saw taxonomies as the enemy - no-one but the indexers and information controllers really understood them enough to get the full use out of them (which is probably still true today). Basically, a lot of people couldn’t find the stuff they needed to get them through their working day.
My take on Matthew’s Grand Unified Theory is that people evolve ways of coping with taxonomies - we are human, we are adaptable, we will find a way. Often the Information Architect can help this process by formalising the existing folk taxonomy coping mechanisms and adding folksonomies.
So, today at least, my position is this: taxonomies are not the enemy - they are vital because they involve the taxonomists in the process - and they allow for at least one browse facet that is common to all people in the organisation. I believe that taxonomies rarely provide a total solution in and of themselves, but that they are at least one part of the total solution.
Humans have been pigeonholing things for a long time. We know this because (archaeological and modern) anthropolists have seen this in cultures across thousands of years. It seems we just can’t help it!
When it comes to the representing of the formlisation of our pigeons into our holes, I think some information professionals have given taxonomies a bad wrap. They try to claim its the resultant taxonomy is best way of representing said knowledge when, in fact, it is only one way of doing so. While in certain circles, it is probably the best way (like Dewy is good in Libraries because Librarians know and understand it), but move taxonomies outside of that circle and it becomes relatively meaningless.
The lesson for IAs IMHO (to carry on with the enemy metaphor) is to ensure taxonomies are used by the guys in the white hats for white hat business and not to hand them over to the black hats.