The feature known as directories (aka "folders") in your operating system is just a pitifully broken implementation of "tagging"

folders are lame. it's like being able to only assign 1 tag to an entity

It's not tagging - it's taxonomy. Which, while similar, is a completely different beast.

Taxonomy is (at least as applied in this context) a broken implementation of tagging as well, although an additional point arguably beyond the scope of my original claim.

For an especially spectacular and well-documented example of why I consider this a broken implementation, please consult ''An Essay towards a Real Character and a Philosophical Language'' (London, 1668) By Bishop John Wilkins; and then review the critique thereof as documented in ''Words and Rules: The Ingredients of Language'' By Steven Pinker (pp 10-13) (an excerpt freely available through google book search)

Please also note that "tagging" and "taxonomy" are not mutually distinct concepts.

To clarify my last point: there is no structural representation within (what you call) "taxonomy" that cannot be implemented within (what I call) "tagging". There *are* however, structures within "tagging" that *cannot* be represented with "folders and shortcuts".

Therefore, "tagging" is a superset of "taxonomy" and the subset is not enough (i.e., it sucks) thus it represents a broken implementation.

<personal_definition>
Tags doesn't have hierarchies. Tags are created on demand. You can use multiple tags for each item.

Taxonomy has hierarchies, and frequently also semantically meaningful categories. They have to be created to be used, and are often limited to one category per item.

Systems that have both these aspects are hybrids.
</personal_definition>

Taxonomy is broken for most rich content (musical styles, what a photograph represents, what subject an essay deals with, etc), while it's useful for things like structuring species.

I'm not sure if directories are that broken for most of the file system. True, turning all of /bin, /sbin, /usr/bin, /usr/sbin, /usr/local/bin, /usr/local/sbin, etc into some kind of bin tag, applying the other distinctions as some kind of tag would be nice - otoh, it would mean I'd have to separate file-called-totem-tagged-bin-and-local-and-not-system from file-called-totem-tagged-bin-and-not-system some other way, but I guess that's just an UI problem.

In my home directory, however, it would actually be really, really useful to have tagging. Which is why (for instance) Microsoft built that into Windows Vista already. Applications like Tracker and Beagle (and I guess Spotlight on Mac) that work with mime types, meta data and tags do represent a better way to access my files. But I'm not convinced it would actually be an improvement to use system wide.

So it actually sounds like you do not dispute my claim, you just apply a different set of definitions. My definition differs from yours only in that what you call a "hybrid" system I call tagging, since a correctly implemented tagging system can make both "taxonomy and tagging" (as you define them) available to the user transparently (since 'taxonomy' is simply a more rigid assumptions of how tags may be assigned and associated with one another).

Since you have yet to dispute the claim that (my definition of) tagging represents a functional superset of what (I call) "folders and shortcuts" (ie taxonomy), I will chalk up our disagreement to a matter of semantics and preferences regarding terminology.

Also, as an aside, even your taxonomy definition falls apart a bit when you consider symbolic links (aka shortcuts in windoze land).

Having said that, I still do recommend a look at the Bishop Wilkins piece though. It gives a stunning example of how even for the seemingly trivial task of classification, taxonomy, can fail miserably.

Cheers,

Directories have a tree structure, whereas tags are flat.

/a/a/a/a/a/a/a/a/a/b/a/a/a/a/a/a/a/a/a/
is a valid directory, distinguishable from
/a/a/a/a/a/a/a/b/a/a/a/a/a/a/a/a/a/a/a/
but with a flat tagging system they would be indistinguishable, 18 a's and 1 b

lrn how filesystems work plz kthnxbye

@James Johnson: kthankxtou2 bbuh

@Rorek: You haven't refuted my claim or comments though, you've just implemented another broken variant of tagging. Who says tagging *has* to be "flat"? Tagging (as I define) allows grouping, nesting, bundling, whatever you want.

(see e.g., del.icio.us / settings / tags / bundle)

usr::local::bin could be a single tag, a compound tag or even a "path" (if you prefer).

Your definition of tagging is not the one commonly in use. Perhaps filesystems could be improved by using a system such as you describe. Feel free to enlighten us on how that would work.

I'm not quite up to the task of writing the specs or proof-of-concept for an entire filesystem, but I do have real-live running code that implements some of this via supplemental metadata and a caching database.

It's not very complicated, the same or similar information architecture has been deployed by others. Some call it by different names, and some systems are now defunct, but I'm not the only one on the Earth who has had these kind of ideas.

For an approach similar (but not identical) to the one I describe here, see http://en.wikipedia.org/wiki/Faceted_classification
and the associated links.

I prefer to simply use the simplifying term "tagging" because "faceted classification" and terms like that just put people to sleep ... unless they happen to be reference librarians or something similar.

Note: I am aware that others would classify this as 'taxonomy' or 'hybrid' and dispute my usage of "tagging" ... nevertheless the point is not the terminology, the point is the prevailing approach is broken.

I think the point is a) your terminology and b) that the prevailing approach serves a different purpose than you imagine.

Use the right words; you're on Jyte.

D'A

i aht ewhen i encounter an interesting but frustrating claim when sleepy. at the moment the most useful coherent thing i can say is "See also: BeOS".