I'd love to see benchmarks testing the two, and out of curiosity also including compressed JSON docs to take into account the impact of payload volume.

Nevertheless, I think there are two major features that differentiate protobuff and fleece, which are:

• fleece is implemented as an appendable data structure, which might open the door to some usages,
• protobuf supports more data types than the ones supported by JSON, which may be a good or bad thing depending on the perspective.

In the end, if the world survived with XML for so long, I'd guess we can live with minor gains just as easily.

You probably wouldn't. The main difference is that protobuf is structured while fleece is unstructured, so you would use it in places where you don't want to (or can't) tie yourself to a schema outright.

It's not serialized from what I understand so it doesn't need parsing. It's sort of a structure+pointer dump. We'll see how well that translates to other languages than C though.