[guyromm]

that one i did take for a spin.

i must say that the experience is quite horrible - that torture of having to write map/reduce functions, added with some erratic behavior in regards to data integrity (inserted entries silently discarded, sync to the remote couchdb instance working somewhat whimsically). as soon as your dataset is sizable in any regard (tens of thousands of records in a collection, if i recall the terminology) it begins to just break apart.

was writing a browser extension, and used pouch with the hope of keeping its persistence local and avoid needing a server. seeing that it leaks tried to trade it for a couchdb server. seeing how bad sync is, and that couch is not very comfortable to work with either ended up throwing the thing in favor of a postgresql+postgrest backend.