Greg, it would be interesting to know what sort of concurrent access to you support? How do you handle concurrent writes to a schema-based collection, when multiple nodes are being updated? Do you support granular locking for concurrent writes to the same collection entry, so that you don't have to chunk you data? Ilya