Dear community, you may understand that currently cjdns and Hyperboria have no real future. With this pace, it won't ever replace the Internet. It merely is a cryptoanarchist programmer playground. 
We have to start low and unseen, we need a mass of applications. And they willingly will build that mass as they are in need of security and easy routing.
Please, read my proposals as they may dramatically change the current state of things.

Global Infrastructural Network System - GINS

- as easy as possible -
Every developer should be able to painlessly integrate their application into the network as it is beneficial for both the former and the latter. Every modern platform should be supported. Android, iOS, Linux, Windows and OS X are absolute requirements and no success may be achieved if any of those platforms is left forgotten. Probably, on some platforms, you will be forced to give up on TUN and emulate its existence. Concentrate on reality. Setup should be as easy as "import" statement in python is. Speaking of which, SDKs for various popular languages and environments should be made and provided to developers.
Setting up should be reduced to hardcoding application-specific key, data reception callbacks and central server address (tracker), if necessary, from which nodes will gather data necessary for operating. Unique address may be generated on the first start. 

- advertise as a tunneling framework -
Lots of applications may require data aggregation, after which it is passed to the listeners, almost unchanged and raw, and this is enough. By the means of p2p connection, these systems may be heavily decentralized, decreasing the main server load, which may then act merely as a telephonist, granting data receivers direct access to the designated data provider:
"A is requiring: Y, grant access for X"
"Y is reporting: yes, sir A"
"A is reporting: X, you can now establish communication with Y"
"X is requesting: Y, give me that sweet data of yours"...

- different applications, one stacked network -
All applications, willingly or not, are placed on the same network, and keep in touch with each other to build a rigid structure, providing every user unprivileged access to bandwidth and routing. Peer discovery over IPv4 should be simplified and automatized, with the use of DHT, either global or application-specific, that will inadvertently, but assuredly, lead to the connection to the global network.

- built over the Internet, to replace it one day -
At first, it will be simply means of connecting different application users, securely, through their NAT's and unknown network topologies. Then there will appear islands of pure physical connectivity, unreliant on the Internet, but still providing unhindered access to it, for those in need. As more and more services and devices join the network, these islands will grow and people will seek connection to our haven, either to use their cjdns-based applications or to connect to the Internet free of charge.

- part of the metropolis infrastructure -
For this to work and start getting out of virtual world into the physical plane, core application set should stay local and target infrastructure, building the backbone for the electronic government. Static local city infrastructure provides fertile ground for our pure-cjdns islands to grow and that's the most realistic way that I can see. Every traffic light and every meteorological sensor may be connected to the network and be its publicly accessible node. 

- mass usage is the only way out -
To exist, network should work, productively. If it is there just satisfy our anarchic desires, it has no use and will die with us and our motivation. Seed the availability, by investing your time into SDK's and frameworks and watch the network grow for everyone to reap its fruits, as more and more people realize benefits of its usage and further development.

Thank you for your attention.