Copied over from #2524 (comment):

@jaekwon @mossid not 100% sure where the KeyPath stuff interfaces with the user. Are we expecting that the KeyPath passed into RequestQuery.Path corresponds to the keys in the ResponseQuery.Proofs? In which case this KeyPath becomes the sort of standard for RequestQuery.Path and should also be documented here?

Noting that it's a bit weird for the general Merkle proof types to live in crypto/merkle. They're effectively part of ABCI and should live in abci/types, just like the ABCI PubKey type...

We would also benefit from separation in crypto/merkle between the SimpleTree code and the GeneralMerkle code. Since we want most of the crypto/merkle to be used by external consumers, it really helps to have separation to simplify looking at the godoc.org.