Mashup Camp 2 included a session on microformats and APIs. While the overall conversation rambled a bit, there were a couple of very intriguing threads. Kevin Lawver of AOL, for example, has spent a long time being frustrated by working with the Google search API. Lawver sees the power of microformats concentrated within the concept of grabbing a chunk of an HTML page and using it as raw data in a mashup. Why create an RSS feed or XSL transport, if microformats can be part of the original HTML markup? Why create new schemas again and again?
Someone has already written a prototype for grabbing chunks of other pages - for example, hCards. There's a Firefox extension called Tails that will convert hCards on web pages into vCards for Outlook. Also, Eventful (FKA EVDB) and Edgeio don't actually collect information directly - rather, they aggregate events from all other sites. (Disclosure note: Omidyar Network is a funder of Eventful.) This implies that a service that generated microformats would be nice to have - whether it was embedded into the original page, or delivered via a middleman.
Although microformats.org has a lot of available definitions, they aren't aimed at APIs. Should they be? Microformats are meant for human-readable data, not for data that is already machine-readable. If HTML becomes a web service language, then microformats obsolete the need for XML. (The DTD is dead. Long live the DTD!)
More info on this topic is available on the Mashup Camp wiki.