This is an aside and also a "you guys rock" to the CouchDB dev team... I watched a talk by Benjamin Anderson at Meteor today: http://www.dataversity.net/archives/6714?t=1320768580
where he talked about their experience with running CouchDB (via Cloudant) for 2 years in production, scaling from what I imagine was 2 or 3 nodes to a 14-node cluster. In the last year alone their traffic has grown 5x while their Couch cluster has only grown 20% to cover that. Benjamin is straight forward with some of the shortcomings they have had, but the take-away from this that floored me and is a testament to what you guys do... in all these years after all these 10s of terabytes of data, they have never lost data or had downtime. They just kept growing their cluster, adding nodes as needed and got back to work. No world-ending events, total cluster failures or total rebuilds. Couch just happily chugged along, keeping track of data and smiling. (not discrediting any world Cloudant put in here to make that happen, I just don't know what it was). He also specifically called out how the backend API to Couch has (practically) not changed at all in the last 2 years, allowing them to focus their development efforts on customer-facing features and not spend time writing and rewriting backend persistence features as Couch grew and released new versions. To all of you committers that spend your nights and weekends on Couch, this is a glowing testament to the hard decisions you have made. Really great job guys. -Riyad