SMS is federated and very difficult to block without disrupting essential services.
See: https://github.com/WhisperSystems/Signal-Android/issues/2818
Of course going the route of using centralized services allows later monetization my Moxie and his brogrammers.
Especially this one:
SMS and MMS are a security disaster. They leak all possible metadata 100% of the time to thousands of cellular carriers worldwide. It's common to think of SMS/MMS as being "offline" or "peer to peer," but the truth is that SMS/MMS messages are still processed by servers--the servers are just controlled by the telcos. We don't want the state-run telcos in Saudi, Iran, Bahrain, Belarus, China, Egypt, Cuba, USA, etc... to have direct access to the metadata of TextSecure users in those countries or anywhere else.
In a way, Moxie's argument (don't tie things to phone numbers, don't use a centralized system for message transport) is exactly why Signal itself is so problematic.
Telcos, on the other hand, are usually based in the country in which they do business (and often have separate businesses in each country, e.g. Vodafone). As we've seen many times, it's pretty easy to get them to cooperate (voluntarily or by law). Much easier to get all the metadata you need that way.
https://whispersystems.org/bigbrother/eastern-virginia-grand...
I'm not going to rely on the US government adhering to its own laws when they've shown they won't do exactly that.
Instead, cryptography is supposed to prevent OWS from even being able to implement a backdoor.
The problem is that generally smartphones (at least on the WAN) do not accept incoming network connections except those blessed by the OS provider (Apple and Google). Said push notifications are tightly controlled and not particularly feasible as a p2p network transport.
I think what would solve this problem is a 2 level p2p system. The supernodes would be raspberry pi's,a plug computers [0], or even a wifi router. They can accept incoming connections, don't pay as much for bandwidth, could be shared among friends/family/neighbors. Even a 64GB microsd + arm would handle substantial messaging traffic.
Then the more network and power constrained phones could check in with their preferred supernode, ideally falling over to other supernodes if necessary. Messaging traffic is tiny, so storing it 3-4 times on redundant supernodes would still be cheap.
Said supernodes would earn reputation with their direct peers in the form of providing bandwidth, storage, and forwarding packets.
That way it would be much harder to censor, fairly robust, and ideally still easy to use. Whoever writes the software could charge an extra $20 for a turn key raspberry pi or plug computer.