Thanks for the feedback. It's not my place to make commitments for the company, but I can tell you an engineer's answer: I've yet to see a logical model for mu

Thanks for the feedback. It's not my place to make commitments for the company, but I can tell you an engineer's answer: I've yet to see a logical model for multi-master replicated documents that beats what Damien chose when he created CouchDB. So I'd be very surprised if that changed, and as long as that stays the same, interop at the transport layer isn't hard to maintain. However, there are lots of ways the transport / wire-protocol that we use to transmit the logical model could be improved. So in my perfect world there would be many alternative transports, and the various couches could implement pluggable handlers for them. In the end this would preserve interoperability but also make room to experiment with the details. If you are interested in those conversations they are happening here: http://mail-archives.apache.org/mod_mbox/couchdb-replication/ http://mail-archives.apache.org/mod_mbox/couchdb-replication...