There is an enlightening discussion going on between Lou Rosenfeld, Clay Shirky and others on the utility of folksonomies as used by Flickr and, vs. subject-matter-expert developed taxonomies. As one of the commenters has pointed out, this is not an “either/or” issue. Certain applications where the scope of the content and users is bounded will benefit from the discipline of a carefully architected vocabulary. Other applications where the scope of either the content or the user community is less well-defined will either suffer or, more likely, the users will ignore the prescriptions (this is why the “semantic web“, if I understand it at all, is hopeless). The key issues are related: cost and adoption (cost is usually a function of adoption, not development), and I think they both would agree on this point. How these approaches might work together is trickier and well worth exploring. In any case, this debate provides a condensed lesson in many issues that most enterprise content managers have probably not thought through, but even those that have should check out this thread.