There’s some additional discussion on the topic of “flat” namespaces going on elsewhere: Wikisonomies and The Chaotic Order
Indeed, I have been thinking of using a system where, to use terminology from these new discussions, my categories are facets to which tags are associated. My current system is one of all tags are categories, and visa versa, with a subtle hierarchy offered by the ability to make some categories parents of other in WordPress.
However, if the categories were facets, and then, perhaps, all tags were stored as keywords I would have the ability to collect posts but also have freeform tags, without the creep of so many categories.
The problem with giving authority over hierarchy to a wiki is that it becomes just another external authoritative schema. The only way to stay with the spirit of tags is to allow users to provide, on the fly and as they will, the relationship between tags. If there were a useful way to specify that a tag has some relationship to other tags …
For example, if one could offer a list in, something like, PREV to tags that are more general terms and, something like, NEXT to tags that are more specific, there would be utility and folksonomic wisdom in that. This would continue to allow the dynamic to determine the relationships, instead of a more static schema. Certainly, this could be reported on programatically, just as tags are now. Imagine a DMOZ like directory that is dynamically rendered from a folksonomic understanding of tag relationships.
This is about how to create a thesaurus. How can a tag be represented as having a relationship to other tags on the fly by the end-user, instead of via a schema? This is not a simple issue. There is a MARC format used to record various classification schemes, for example, MARC Classification which addresses the issue of Tracings and References.
Relationships include See From, See Instead, See Also, Broader Topic, Narrower Topic …
My preference, when thinking about implementing classification schema is to offer a thesaurus of relationships, but also include a freeform field in which a user can characterize the relationship.
Perhaps the folksonomic way of handing this is to offer only the freeform. A field which offers to complete a sentence between two tags could be very useful. For example, “X is similar to Y” where X and Y are two otherwise arbitrary tags and “is similar to” becomes the statement of relationship.
Another option would be simply to artificially promote facets as special tags which contain other tags. Meaning that one simply takes some tags to be more important than the others, an artificial two-layer hierarchy, but ultimately that seems to be just the beginning of a slippery slide.
Why not let go of defining the thesaurus in the same way that tags have been liberated?
Every author’s domain of tags is potentially distinct from all others, though functionally there are more convergences that divergences, and that domain may shift over time for even the same author, as tags develop new and altered meanings to the author over time. So, to extend this logic, each author has the potential to have their own, and should be assumed to have, their own classification scheme, and perhaps many schema, especially over time.
Update: Here’s a link to an implementation of a radial browser that allows one to explore data in the CIA worldbook. The more important element is that the radial lines have smaller spheres which express the nature of the relationship between the nodes. This is an interesting programatic way of representing classification data. This is a really simple, elegant demonstration.