Labnotes at Opsop riffs on my recent post on Open Tags, and suggests using UML "anchors" or "fragments" to denote tags:

[from More on Microformats]

I've also been thinking about using the fragment identifier, e.g. <a href="#thai" rel="tag">Thai</a>. It's semantically correct (I know what this tag means to me), and persistent (back reference), but I'm not sure how it plays along with syndication.

None of that would matter if I didn't have to link to anything. When I write my post, I simply add a line of text that looks like:

tags: microformats reltag semantics xri

That's enough content for people who read the post, it's also enough content for the WP plugin to figure out the tags and create the microformat. I only need to add the links so Technorati can spot the tags. Understood. That's what the "rel" attribute does. But then why does relTag have to say anything about the URL?

Why, indeed?

I think there is some merit in the simplicity of the use of an anchor ("#thai") but it poses other small problems: don't hey have to be unique in the HTML file? Or maybe on the the first instance would be found? But of course, the way that we use tags now -- a pointer from a taggregator like Technorati points to a post, not a specific location in a post -- doesn't gibe with how anchors work.

Even more simple is the convention of a keyword "tag:" but that brings up a host of problems, because that sequence of characters is not encoded in any way, and people could use it for other purposes, like "I went to look at the tag: it was red" does not really introduce three tags: "it" "was" "red". Also, there is no obvious way to end the sequence.

No, I really think we do need something like the XMI approach that Drummond proposes (see earlier post), but I am not certain yet of the exact form it should take.