IDEAS

one issue is compound queries like table('a').union(table('b'))

i was thinking of doing a quick scanning the json on the way out for [15, table term (or hooking the first .table() term method call).... then send it to that server that's in charge of the table ... i would think it could be a quick optimization .... in the case above , i would have thought just sending it to the server in charge of table A would be good enough without the driver fully "understanding" the whole idea query processing..... was mostly trying to avoid sending the query to a server C that wasn't in charge of any table A or table B.... there was a post about how someone used proxy mode and it totally reduced DB server load by a lot.

and if there's no explicit table term, send it to any server in the cluster