Why websockets and HTTPS? That's way too heavy for embedded devices, and why do you care what the transport is if the messages themselves are encrypted and authenticated?
You are correct. If the transport is secure, then the protocol doesn't matter. I just think that a central controller model could use a protocol which is more suited to keeping the code on the mobile device as simple as possible. If you just want to send MQTT commands to nodes without a central controller, then what you propose makes sense.
The problem is that TLS is far from "as simple as possible", and you need TLS on the device in order to have encrypted communications just device-to-server.
In my opinion, device-to-device encryption is much better, and you don't need all the overhead of TLS. I've actually finished a first draft of the prototype and am looking for people to kick the tires and discuss it. Would you be interested?
It would be reimplemented in the target language (or possibly C). The interesting part is mostly the protocol, the language is just the proof of concept for now.
Comments
Why websockets and HTTPS? That's way too heavy for embedded devices, and why do you care what the transport is if the messages themselves are encrypted and authenticated?
If you want to take a look at what I'm building, I'll publish it at https://github.com/skorokithakis/stringphone when I have some code.
You are correct. If the transport is secure, then the protocol doesn't matter. I just think that a central controller model could use a protocol which is more suited to keeping the code on the mobile device as simple as possible. If you just want to send MQTT commands to nodes without a central controller, then what you propose makes sense.
The problem is that TLS is far from "as simple as possible", and you need TLS on the device in order to have encrypted communications just device-to-server.
In my opinion, device-to-device encryption is much better, and you don't need all the overhead of TLS. I've actually finished a first draft of the prototype and am looking for people to kick the tires and discuss it. Would you be interested?
If you make it open to everyone, I'll take a look.
Post the code on Github and document what it does in the README.md file.
Done.
How would you run this on a mobile device? Since it's written in Python, you'd have to run it with something like Kivy.
It would be reimplemented in the target language (or possibly C). The interesting part is mostly the protocol, the language is just the proof of concept for now.