Excellent suggestion, and something we're already considering for a future release. Our current reasoning is that if you are only pulling back one or two objec

Excellent suggestion, and something we're already considering for a future release. Our current reasoning is that if you are only pulling back one or two objects, then an extra round trip probably isn't that bad. If you're pulling back many objects for processing, then you might be better off feeding the query results into a MapReduce job so that it runs in parallel across the cluster.