|
I agree, but go farther: Most things have multiple kinds of interesting properties. And in general, the more complex the thing, the more interesting properties it has. Ofc "interesting" is relative to the user/observer. The problem with hierarchical taxonomies, and with taxonomies in general, is that they try to categorize things by a single property. Not only that, the selection of the property to classify against, is relevant to the person who made the selection, but it might not be relevant, or at least the most relevant, property for others who need to categorize the same set of things. Sometimes people discover "new" properties of things, such as when a new tool or technique for examining the things, comes into existence. And new reasons for classifying come into existence all the time. So a hierarchical taxonomy begins to become less relevant, as soon as it is invented. Sometimes one wants to invent a new thing and needs to integrate it into an existing taxonomy. But they have a new value for the property that the taxonomy uses for classification. Think back to SNMP and MIBs and OIDs. Now the original classifier is a gatekeeper and you're at their mercy to make space for your thing in the taxonomy. In my experience, the best way to classify things, ESPECIALLY man-made things, is to allow them to be freely tagged with zero or more tags (or if you're a stickler, one or more tags). And don't exert control over the tags, or exert as little control as you can get away with. This allows multiple organic taxonomies to be applied to the same set of things, and adapts well to supporting new use cases or not-previously-considered use cases. |