"Some use tags as folders to house objects, others use them as descriptions of objects."
Well put, and I would venture to add "tags as groups", as in the object being a member of a group.
Still semantics, dependent on where you come from, where your head's at that precise moment.
But all good and useful angles I would add.
[NOTE: My following points are dependent on the use of multiple-tags-interception method / anataxonomy-in-practice (pls use Firefox/Safari) along the lines of this example to find an object / subject.]
Tag as description:
- Adds meaningful knowledge to the namespace, in particular if we had some information about who tagged, when it was tagged and why the tag was chosen (source analysis in history comes to mind).
- Pro: Can be rather precise.
- Con: Too precise and may require assimilating logic of tagger.
Tag as membership to a group:
- Close to "tags as description", following the classic human attitude of "Show me your friends and I'll tell you who you are" kind of approach. Useful, but I would have to keep those tags quite dynamic, first impressions having a tendency to be wrong :)
- Pro: Almost like a metatag, gives lots of information in one go. Imprecise, requires less logic assimilation.
- Con: Very wide, very metataggish and have thus to be quite dynamic as view of object/subject changes over time.
Tag as position:
- Useful to describe physical whereabouts of course, easy to fall back to when in a tree-structure-mode.
- Pro: Good starting point when moving from tree-structure, porting my 593,406 files from tree-structure to tag based structure could be by tagging purely with folders, including sequence of course. Then add meaningful tags as time passes.
- Con: It's after all tree-structural and sequence matters thus assimilating tagger-logic is required!
It does not matter where we start as long as we let the process flow towards what is natural, whatever that would be (you know where I stand, but this works even if I'm wrong - cool eh?).
No extra and precise tag will ruin for the other tags. No imprecise tag is useless unless on its own. Even tree-structural tags can be useful as it tells a story of times past :)
Add the paradox raised by Tom that a hundred year old subject/object where "obvious" tags would change over time: Using the "anataxonomy in practice" the new or old tags would not matter, modern man would use the current cloud while the ultra-conservative would use the old set - and none would ruin for the other!
Example: Take a HR system using tags only to "organise" people, start with a virtual hierarchy with departments and positions as tags - and the current managers will find the person. Then add tags to add knowledge about the people, factual descriptions (speaks fluent Italian), subjective impressions (grumpy in the morning - tagged by subordinate, friendly in the morning - tagged by boss ;)
Add more tags - at some point the hierarchical tags does not matter any more. That would make the transfer from hierarchy to tag based structure rather painless!
Knowledge of the person would move from the "heard in the cafeteria" or "fifth paragraph, second page on a CV in some drawer" to a common space adding transparency, better use of resources and dynamic increase in namespace-value.