A couple of hours after I gave my talk about Flickr machine tags and their possible lessons for Twitter’s new annotations, Raffi Krikorian gave a talk at Warblecamp on that very subject. He’s now posted slides of the talk, which are well worth a look.
In them, he expands on the format for annotations (they consist of types, attributes and values; types can be repeated, but attributes can’t), and mentions an annotations “explorer”, which will contain both “statistics of most used, adopted and trending attributions” and a “wiki page so developers can document their attributes”.
This dual approach pretty much fixes the main points I was worried about, combining a “pave the cowpath” method (looking at actual usage data) with a more editorial take on the wiki.
Anyway, the talk touched on even more (including the beta rollout plan, which will be based on OAuth-enabled apps, rather than feature flags or user lists), and mentioned release dates (which are reassuringly close). All in all, it’s pretty exciting, and I’m looking forward to seeing how they get used in the wild.