I've been looking at the source, but not quite find the things I'm looking for, so I have a few questions. Are columns for a row stored in a serialized data structure on disk or stored individually and put into a data structure when the call is being made? Because of the slice query, does that mean that all columns have to be read in before any are being sent back? If that is the case, could it be more efficient to use rows instead of columns for storing for example indexes and you just want to get a few at a time?
-- Regards Erik