So, I have completed the first draft of the ResultsDB 2.0 API.
The documentation lives here: http://docs.resultsdb20.apiary.io/#
and I'd be glad if you could have a look at it.
The overall idea is still not changed - ResultsDB should be a "dumb" results store, that knows next to nothing (if not nothing at all) about the semantics/meaning of the data stored, and this should be applied in the consumer. This is why, for example, no result override is planned, although it might make sense to override a known fail to pas for some usecase (like gating), it might not be the right thing to do for some other tool in the pipeline, thus the override needs to happen at the consumer side.
What's not covered in detail is auth model - I only reflected it by acknowledging the probable future presence of some kind of auth in the POST queries (reserved _auth parameter), but the actual implementation is not a problem to solve today.
On top of that I'd also want to know (and this is probably mostly question for Ralph), whether it makes sense to try and keep both the old and new API up for some time. It should not be that complicated to do, I'd just rather not spend too much time on it, as changing the consumers (bodhi, as far as I know) is most probably much less time consuming than keeping the old API running. At the moment, I will probably make it happen, but if we agree it's not worth the time...
Feel free to post comments/feature requests/whatever - I'd love for this to be stable (or at least a base for non-breaking changes) for at least next few years (lol I know, right...), so let's do it right :)