I think Shirky’s got hold of half the truth. It’s an exciting half, and high-speed computers mean that we can use folksonomies in a way that approximates the ultimate folksonomy that is language itself. But once the folksonomy of language becomes rich enough, and once literacy enters the picture, language also becomes a taxonomy (as it were) and we begin to teach aspects of it prescriptively as well as descriptively.

The other reason is related, but not so rigorous. In doing research, a stable set of tags is helpful. This we already have lots of experience with, actually, and I try every semester to get my first-year composition students to use it. Each database the library subscribes to tags its entries with different words (i.e., descriptors, or subject terms, or keywords, etc.). For the same thing, different databases use different tags. A search (folksonomic style) will get a start on research, but at some point using descriptors (taxonomic style) is much more effective. I expect the same will be true on the web.

Fun tid-bit–one of the competitors for the name RSS was “RDF Site Summary” — RDF being a rigorous, stable-meaning set of metainformation. That’s lost out (I think) to “Really Simple Syndication”. I’m fascinated by the switch from “Summary” to “Syndication”. The implicit or explicit function of the content might guide the use of taxonomic or folksonomic tagging.