Having the network boundaries be explicitly nonprivate and giving users the ability control when/where/how they negotiate the private/nonprivate boundary seems

Having the network boundaries be explicitly nonprivate and giving users the ability control when/where/how they negotiate the private/nonprivate boundary seems like a reasonable design choice to me. It makes the network more interoperable with nonprivate networks. If the whole network private, then the privacy faults can only happen at its boundaries--which are places that the protocol designers have less control over. The alternative is having to wonder about what kind of identifying metadata the exchanges are leaking--and they're an easier target for an adversary. Also, I have to imagine that the everywhere-private nature of XMR is why I don't see it on my exchange's list, while I do see ZEC there. But I wasn't talking about that. With zero-knowledge proofs protecting the shielded transactions, your anonymity pool is essentially the entire set of people that use shielded transactions. With the Monero approach, your anonymity pool is large, but it's still a subset of the whole network.