Technical considerations.
Re-inventing everything is too much. It will be enough to have a new version of the very lowest layer of the protocol. That means a new version of IP. IP is unreliable and insecure and yet the entire Internet runs atop of it. Functionality similar to DNS will also be needed.
It will need to scale out to billions of devices.
It will need to have a hierarchical architecture and local routing should be handled locally then regionally, then nationally then globally.
Concrete implementation ideas.
Since The Internet was created computing power has increased more than 2 orders of magnitude. It would be sensible to allow that we might have an implementation that uses more computing and bandwidth. I4T addresses should be unlimited.
Premature optimizations should be avoided while we concentrate on the core functionality. Eg. Local gateways will eventually reduce the connection cost and Ipv6 routing will reduce the bandwidth costs.
It can run on top of the existing Internet.
Cost
It’s not free but it’s very very inexpensive. I estimate that a thermostat or smart switch can utilize The I4T for $0.001 per month or $0.012 per year. Moreover the bulk of that cost is for maintaining a connection. Whenever a hub or gateway can be used the cost is only for routing and name resolution and is 100 times smaller. Since a non-profit is funneling all fees paid to pay for computing one could imagine that The I4T will be less expensive than any other conceivable option.
Conclusions
While my meager contributions here may or may not prove useful it’s clear that eventually The I4T will arrive and will have these desirable properties.
About me
I’ve written a version at https://github.com/awootton/knotfreeiot which is running live now. Your comments and contributions would be greatly appreciated and noted.