I'm all for the idea of a long range, decentralized, encrypted mesh net. I've read a fair number of papers proposing and analyzing different approaches to make this happen.
Unfortunately, unlike the internet, or a local mesh network the total collective bandwidth isn't there. If it worked people would try to do full duplex audio with loved ones, file sharing, monitoring GPS coordinates of fleets of vehicles, internet gateways, and a zillion other bandwidth intensive things.
Bandwidth is so low, that even sending bytes is rare, encoding that average a few bits per character are common. Such encodings try to minimize the bits by using less bits for more common english characters, so they often are measured in words per minute instead of bits or bytes.
There was exist various hybrid solutions like APRS (which can use ham + internet), winmail (ham + internet email), various digital radio protocols (ham to tower, which has an internet connection). But a pure radio mesh net just doesn't seem practical.