Even though he was completely wrong about how the Dinnerbuzz solution works, Jeff Jarvis is dead-on in his Made For A Distributed World. Jeff thought that Dinnerbuzz would aggregate blog posts tagged with various sort of restaurant information -- "Thai" "Reston" "Virginia" -- and then provide a search mechanism so that people could find posts using the domain schema for restaurants: "find all Thai restaurant reviews for Reston Virginia". Alas, it turns out to be not constructed that way.
Jeff's point is that in a distributed world people want to post all their observations at their own blo, not over hundreds of thematic blogs. I would like to post about wine, restaurants, music, and gizmos at my A Working Model blog, and simply ping (via trackback, perhaps) other aggregators (like the idealized Dinnerbuzz site Jeff was dreaming of), and they could sort out what's what based on the tags.
Jeff suggests that either Technorati or Delicious style tags would work, but an expanding world of services mining meaning from tags mean that we would start to have dozens of tagspaces. Currently, a technorati tag for "Thai" looks like this:
<a href="http://technorati.com/tags/thai" rel="tag">Thai</a>
Which means that the tag is managed by technorati at that URL. I would like to shift over to distributed tags, based on a relative path in the href:
<a href="/tags/thai" rel="tag">Thai</a>
This would mean that the dozens of sites I might be distributing my reviews to could all inspect my tags, and detemrine if they were relevant for thier use. As Jeff points out, we could include specific tags as guidance, like this:
<a href="/tags/for:dinnerbuzz" rel="tag">for:dinnerbuzz</a>
This would mean we could reuse the same bunch of tags in a single post for wine, cigar, and restaurant reviews, for example. More importantly, we could write posts and tag it independently of the service or services that will ultimately reuse the information. We wouldn't be locked into a specific service (like Technorati or Delicious) and we could begin to operate in an open tagging model, which we are going to need.
[Note: I tried to dig around in this topic a few weeks ago, in a discussion with Dave Sifry of Technorati in which I was referring to "remote tags". But this idea of Open Tags is a much cleaner treatment.]