Running a bit further along the ideas in this posting about revealing a small programming language over your web site’s data.

Today I’m thinking that this is an interesting variation on the idea of a data exchange standard; something much richer.

Consider this URL that assembles a page from the union of URI tagged with a single token at both del.icio.us and at flickr.

     http://oddiophile.com/taggregator/index.php?tag=market (try it)

Or the same idea as executed by Technorati.

     http://www.technorati.com/tag/market (try it)

In both cases the ceiling on what’ possible isn’t very high because the query language available is limited to “What URI does site FOO have with tag BAR.” Clearly a richer query language is desirable. One that let you make queries like: “What URI does site S1 and S2 have with tags T1, T2, but not T3 that site S4 doesn’t tagged with T7 excluding any tags made by the group untrusted actors or commercial actors.” Rich queries demands more consensus about the data model and the operations upon it.

The current bloom of fun illustrated by the two examples above stand on a a very small consensus, i.e. “Yeah let’s mark up URI with single token tags!” A slightly larger consensus would enable a larger bloom.

A small consensus, like the tags, is a lot easier to achieve. It is less likely to fall victim to the IP tar pit. It’s maximally likely to be easy for the N sites to adopt and rendezvous around. It allows one site to set an example (as I’d argue del.icio.us did with tagging) so that the other sites can mimic the behavior.

I wonder if the idea of a simple query language along the lines of the one my earlier posting could enable that. Interesting design problem.