Getting a little too sharp in writing the PyCon paper (egads, it's something like 3.5 weeks away now, we should get a hotel some day soon). It's too easy to start pointing out the little flaws and problems in the conceptualisation of the two protocols (data and non-data descriptors). I don't want the paper to sound too critical, after all, I use and love the features, and the little figure flaws are mostly due to the evolution of the features over time.