Free networks

Free networks

Did you ever get the idea that one could use a computer network for free if every network participant ran his own set of hardware, and these hardwares were mutually linked in a wireless way through the air?

Packetloss

With mass-available radio networks this is not practically possible because a radio network has typically elevated packetloss (1% is not uncommon). Chaining links in such a network would result in a fast increase of the packetloss above an acceptable level and all TCP would grind down on such a network.

But with Ronja, which has a typical BER of 10^-9, this is easily possible. Even chaining 1000 Ronja links would still keep the BER at level of 10^-6 which is still usable for TCP connections.

Full duplex

Another problem of a radio network is that it is halfduplex. The difference between half duplex and full duplex is like between a stacked freeway interchange and a level crossing with traffic lights. The maximum channel capacity available for one direction is then not constant in time, but varies with the amount of traffic in the other direction. Moreover, due to the used CSMA/CA access method, as the channel is filled, the roundtrip jitter goes up. These are conditions that make a radio network consisting of many chained hops very hostile to TCP connections. My practical observation of existing larger-scale radio community networks confirms these predictions - the network is usable over few hops, but farther the throughput goes drastically down and TCP connections often grind down and need to be restarted.

Therefore if someone tries to build a free network using radio devices, the user will get only very limited value from the network - reliable connectivity only over a limited number of hops. The network would be prone to congestion, too. I believe this is why community networks generally have problems developing into something seriously usable.

Theoretical possibility of a seriously usable free network

A community network based on Ronja would bring much more usability - one could connect reliably anywhere in the network, because the packetloss would stay low and the latency very low. If some link of the network would be overloaded, the bandwidth would be divided fairly between the participating flows, because from TCP point of view, a Ronja link is like ordinary wired Ethernet.

If users would be satisfied in this way, they would have a greater incentive to participate in the network and invest their hardware into maintaining their own node. It would be then just an organization matter to ensure that the users maintain their routers in a reliable state and the network access is divided in a fair way.

Why are we paying for Internet access when at least theoretically, a completely free network is possible at least on the metropolitan area level?