Protobuf is appropriate when message size is the overriding concern. One does need to understand the tradeoffs it demands in return. It is complex and uses different compression strategies for different data types. The byte stream that results from its compression algorithms is totally unintelligible and unusable without a the schema used to compress it. This requires an additional distributed schema registry to function which, in turn, presents versioning difficulties when data structures evolve and change.

Sometimes an application requires extreme compression and sometimes it does not. Sometimes a self-describing message format with lower compression is more appropriate because it minimizes versioning complexity and manages change better.