Write QDataSets that vary over at most 3 independent variables as a Das2 stream. This is the general structure of arrays streamed in a das2.3 stream x [yset yset ...] [y y ...] [zset zset ...] [z z ...] [wset wset ...] [w w ...] Though it's unlikely that all possible components will ever be needed to describe a single dataset. All binary output is streamed in machine-native order. Since datasets written on one architecture are most likely to be read on the same architecture this choice causes the least amount of byte swapping. This is a direct serialization of QDataSet and does not require any legacy das2 classes