One of the things that I've tried to point out about "tags" is how they are a more constrained form of relating information than what is offered by HTML links on the web (and, likewise, more constrained than the various forms of linking planned in the iCite net).

Specifically, tags are dependent on tree structures.

So, I've been at home sick all day today yesterday, and (given the mental clarity one gains when delirious with fever) I've decided to make a few diagrams that highlight how tags are dependent on tree structures, i.e., how tags function as part of (and reinforce) a stricter hierarchy than web links in general.

This first diagram represents a summary of the hierarchy of Flickr in which Flickr's tags appear.

(Note: All diagrams shown here click through to larger images.)

And, this second diagram represents more detail of how individuals' photos and any associated tags live within a relatively elaborate tree structure.

The black lines all represent stricter parent-child relationships within a tree, and the colored lines represent the relational structure between tags and photos.

So, the point of these diagrams is not to say that tags are trees, or to suggest that individual tags function in a strict parent-child relationship with what they link to. It's just to point out how "tags", as a whole, function in a manner that is not independent of hierarchies—they are children within relatively strict parent-child relationships of a tree structure.

So, in summary: Each system of tags lives within a tree structure. These tag systems do not work independent of the these tree structures.

In contrast, hypertext links between web pages can be used independent of tree structures. This third diagram represents a relational network of links between web pages.

What's particularly important to recognize about this distinction between tags and hypertext links in general is how tags have a conceptual "parent", e.g., we have Flickr tags — Flickr (parent) tags (children). Links between web pages can be made without defining parent-child relationship.

Tags assert trees, rather than oppose or eliminate them. Tag systems utilize a parent context to give authority to the tags, i.e., the tags work because of their hierarchical dependence on that authoritative parent.

In a truly relational context of links, there is often no definitive parent, and there is likewise often no definitive children. With web pages, we've had and still have the means to link without trees. But, given a mostly tree-less dialect, we've actually put a lot of effort into developing tree-expressing idioms (more on this soon ;-).

***

So, I say all this because I think what's great about tags is not how they subvert hierarchies—they do so less than hypertext links in general. If you want to subvert hierarchies, think: hypertext links!

What's great about tags is how they balance hierarchy and relationship.

Flickr's tag system is great, and it's great significantly because of how the tag system utilizes the Flickr tree structure—which is also well done. Likewise, the Flickr tree structure benefits greatly from the tag system.

The lesson of tags is not the replacement of tree structures, nor the replacement of hierarchies. I think the lesson of tags is in the example of using hierarchical and relational structures in a way that brings a necessary balance between authority and independence.