Meh. I can't take this kind of criticism seriously. The token format is the *least important* thing about the protocol but it's what everyone sees and comments on because it's on the very surface.

It was a very good choice so far. We have introduced spending conditions, witnesses, DLEQ proofs and more without having to change the token format because we're using base64 JSONs. What matters is that we have well-defined models.

Right now, there is no reason at all to optimize prematurely (except for QR codes). Eventually we'll use a binary encoding once the token format stabilizes – but right now, it's not in the priority list.