I'm coming at this a little late for the blogosphere, but it's still an interesting topic. This spring, I read David Weinberger's Release 1.0 column on Taxonomies to Tags: From Trees to Piles of Leaves (or see an extended abstract at his website). And I just read Clay Shirky's similar article, Ontology is Overrated: Categories, Links, and Tags. Weinberger is also working on a new book, Everything is Miscellaneous, and he has been talking about this quite a bit on his blog (see Technorati for this tag to find his postings on the book).
Both of these articles have to do with comparing the traditional (formal) world of categorization with the new (informal) ideas of tagging and group categorization. Shirky's article is higher in my consciousness, and he talks about why and how categorization schemes developed: primarily to figure out where to place physical objects in limited physical space. In libraries, a book can only be in one location. In museums, artwork can only hang on one wall, or rock specimens can only be located in one place. In the digital world, this limitation goes away, since one can put many labels on a thing.
Another problem with formal categorization schemes is that schemes do not easily transfer from culture to culture. And even within a given culture, the scheme needs to change over time as the culture understands things differently. Again, with physical objects, it is very difficult to decide that entire shelves of books need to be rearranged, based on new cultural norms. And again, in the digital world, this becomes so easy that the constraint of changing the scheme vanishes.
What Shirky argues is that formal categorization itself is unneeded, where "formal" is work done by a set of experts for the good of everyone else. Why not let the people who use the materials place labels on those materials for themselves? With massive numbers of people labeling for their own purposes, and then sharing their labels with everyone else, you will get enough variety of labels in order to track things down. Sure, you won't get a perfect labeling mechanism, but Shirky argues that this is actually a good thing because those terms mean different things to different people.
The Weinberger article covers similar ground, but it goes into more detail on the types of things one can do with taxonomies (trees) and tagging ("leaf piles") and something in the middle that leads to faceted classification. Weinberger also highlights a number of companies that are working in the taxonomy and tagging spaces.
The thing I still want is something like what Siderian have done with fac.etio.us, that lets you slice tags multiple ways. So you can look for reference tips on writing by exploring items that have all three tags. I want this for personal use as well as to browse folksonomies like fac.etio.us does with del.icio.us content. (Unfortunately, it seems to be only surfing the most recent four months of bookmarks.)