This looks great to reuse protobuf definitions in a high performance situation. I would be curious how the performance of this compares to flatbuffers given a reader or byte slice as the use cases seem similar.