I don't agree with Clay Shirky's article Ontology is Overrated: Categories, Links, and Tags.
In fact, what he founds to be shortcomings of ontologies are not only facts known since the foundation of the discipline itself (Aristotle, circa 300 BCE), but in fact the driving motivations of the study of knowledge representation.
Of course everyone's view of the world is subject to biases and prejudices. Biases are what makes knowledge possible (see Kant). (Shannon's) information is always "information about a difference". You have a difference when you notice that something is different than you expected. But... if you "expect" something, you have a bias, and a person with different biases sees different things. Categorizations are meaningful only because they are biased. Successful "knowledge representation" (think of mathematical and formal models, theories, etc.) are always highly biased: they are abstractions that throw away all the stuff that is irrelevant for the intended purpose.

By the way, in my work as a developer of software tools for knowledge management, I found that my customers are always highly biased, aware of their biases, and they want a highly biased information system - which is perfectly functional for them. Notice the different perpective: for Shirky biases are laughable defects (he makes fun of soviet catalogs based on marxism) whereas in my opinion biases are where knowledge is. No bias, no knowledge (this is an very old idea in philosophy).

Here comes the interesing point. The goal of ontologies is to make biases (or "definitions", or "intended meaning") explicit. To formally define them, in order to enable "communication". By the way, the problems outlined in the article are not only restricted to ontologies: they are, more radically, related to what linguists call "the illusion of communication". I'm writing this text "hoping" that we share the intented meaning of these words, and so you can understand what I'm saying; this is true only to a certain extent.
This perspective raises a lot of challenges, some of them are identified in the article, and they are part of the research programs dealing with formal ontologies.

By the way, for many KR/KM-related uses, Folksonomies are not the right tool. I'm not interested in having a thousand people tag a document as related to the topic X, if they do not agree on what X is. Ok, of course this might be just fine for del.icio.us, but if you want to trasfer your electronic clinical folder from Rome to Tokyo for performing surgery, you probabily would want to be sure that the differences between the two information systems are well understood at the meta-level... which is in fact the ontological level.