I'm a little biased, being a pretty heavy CouchDB user, but it seems like Mongo is claiming that their auto-sharding is the defining reason to use Mongo. The forthcoming release of Couchbase 2.0 is going to have the same type of functionality, with buckets distributed across the cluster rather than required replication of the full dataset. Couchbase 3.0 should (if I remember correctly) have the functionality for both the key/value store and for view results.