SubStack: how's the overhead? are you encoding the binary data at all?