It turns out we can easily support schema evolution in Kafka/Avro by specifying a default value for each field in the schema created by the SAL/Kafka producers.

This is a simple change. The main question is whether the default float value should be 0 (like DDS) or NaN. I doubt it makes much difference, because by the time the SAL/Kafka producer sees the SAL data, all fields that are in the current schema have been filled in. Thus this default only affects fields that we remove from the XML. I am inclined to use 0 because it appears NaN support in Avro is somewhat iffy and it matches OpenSplice.