Ben Hyde has been on fire recently. His latest musings regarding URL micro languages and rich, beyond keyword, tag query languages speaks to what the tagged up future might look like. I'm sure somewhere the DB community has explored much of the potential query language design space, especially in the context of SQL. However, a lightweight, user friendly, Web query language would be interesting. Maybe a reduced or tweaked XPath could fit the bill.
Also, I'm in the same camp as Hyde in that I really use this space more as a lab notebook then as a vanity press. As I tell anyone who'll listen, the data model of timestamped posts can be used for a number of different applications, which is one of the major conceptual wins of blogging. There's some structure, but the model isn't so rigid you can't bend it to other purposes.