Geomesh is a Layer 3 mesh network protocol, intended for efficient routing of packets through a network without a central authority, while allowing any node in the network to choose their own address.
GeoMesh intents to address the scalability or efficiency problems that other meshnets have. Traditionally, packets are routed using XOR distance between node addresses, where a packet woud be greedily routed from one node to another based on whichever neighbour is closest in terms of XOR distance.
This works, and eventually gets the message where it needs to go in O(log(n)) hops through the network (where n is the number of nodes).
However, there is no useful relation between the position of a node in XOR space, and the position of that node in the real world. A message that takes a short route through XOR space might go all the way through Timbuktu before reaching the house across the street. Some networks use tricks to cut down on this behaviour, like reducing the distance between nodes, but the fundamental problem remains that the path through the network is somewhat arbitrary.
GeoMesh aims to solve this by using the coordinates of the location of each node (or an approximation thereof) to allow nodes to send packets in the right direction without forcing them to know exactly where every single recipient node is.
To create a network that focuses on:
-
Efficient routing: get a message where it needs to go while avoiding excessive detours.
-
Goal-centeredness: The primary goal of this system is to route packets from A to B. If it does something else at some point, this must be so as to improve routing.
-
Scalability: avoid causing the router to explode once the network gets big enough.
-
Openness: Anyone is allowed on the network. Find a nearby peer, and you're in.
-
Freedom: All nodes should be free to exchange information, as long as they do not prevent other nodes from doing the same.
-
Open source: The more eyes on the codebase, the more mistakes are spotted and the more ideas are applied. It is also GPL-licenced, so anyone is free to create to fork the code should they wish. However, especially in the early stages of the network, sticking to the reference implementation is probably best for now.
-
Re-use of as much as possible of what is already there: the routing software should work on existing routing hardware, if possible even on low-end hardware with limited computing power and memory. It should avoid touching other layers than layer 3 of the OSI model as much as possible. Requiring the installation of special software on the nodes is permitted, but it should be possible to bridge the network over other existing networks.
-
Automation and ease of use: The routing software should be usable without requiring extensive technical knowledge. Good design of the administration interface will be important, and should especially focus on the seperation of simpler and more "advanced" features. Documentation will be important, both technical as well as step-by-step guides.
-
Clean code and quality: A clean code base with proper modularisation and abstraction are important.
Every packet in the system will carry more or less accurate location coordinates of the recipient, which is bad for anonymity.
However, anonymity is not a goal for GeoMesh, as this would distract from the main purpose of routing packets.
If you do want anonymity, use a tool meant for the purpose like a proxy or some anonymity network like The Onion Router (TOR). There is nothing in the protocol that prevents a system like TOR from running (even partially) on top of GeoMesh, but nothing forces it onto those who don't want it.
Once again, getting information where it needs to go is the primary goal, and security on the network will only focus on said primary goal. For example, the network should be resistant against malfunctioning or malicious nodes that prevent correct packet routing or cause excessive unneccesary load.
Use tools like SSL or SSH if you want to send any kind of sensitive information. This is encouraged, even, but not appropriate as a fundamental part of GeoMesh.
Routing works quite well in most cases, even when the graph is not perfectly planar. It still has trouble with certain shapes in the network, but those shouldn't be too much of a problem with a few additions to the protocol. Detecting those strange shapes is all that's needed, the greedy algorithm itself doesn't need to change.
The first pings have already been sent and received over a direct connection between two hosts, albeit very slow ones. The DHT for location lookup is currently being implemented, and the web interface is planned as next on the agenda once we get a simple node-to-node interface going.
It's mainly based on greedy routing, with face routing as a backup. This guarantees delivery in planar graphs.
However, the plan is to use face routing mostly as a backup, with the greedy algorithm being used 99% by letting the nodes talk to eachother to auto-discover network topology, nd make smart guesses as to which nodes will get a message closer to where it needs to go.
You may have noticed the "Openness" bit in the project ideals.
At the moment, there isn't really a network to speak of, it only runs in a simulator and occasionally on simple direct node-to-node connections. The code is there, though, and it's GPL-licenced, so grab a copy!
By the way, if you manage to get it to compile on your system, please tell me!
-
(Virtual Coordinates (Not yet implemented)): A. Rao et al, Geographic Routing without Location Information, University of California - Berkeley Retrieved from https://www.eecs.berkeley.edu/~sylvia/papers/p327-rao.pdf on 23rd January 2016
-
(GOAFR+ routing (NYI)) F. Kuhn et al., Geometric Ad-Hoc Routing: Of Theory and Practice*, http://www.inf.usi.ch/faculty/kuhn/publications/podc03b.pdf
-
(Non-planar routing) http://vega.cs.kent.edu/~mikhail/Research/voids.pdf