Would avro make more sense in a big data world? As I understand it, avro has all the same capabilities of protobuff but it is much more ubiquotous in the big data world, being natively supported by all the big frameworks (spark, flink)