Networking overview

qaul.net implements a heterogenious, fully decentralised mesh network. What this means is that transmission channels can change between device links: the network is made up of roaming network shards. To allow routing across network boundries, qaul.net uses a public key address space, to avoid having to deal with network collisions, and NAT.

For links that rely on regular IP spaces (v4 or v6), a local lookup table is kept in each node that operates an overlay endpoint.

This way qaul.net can run on unpriviledged devices, because it doesn’t rely on modifying kernel routing table parameters to create a DHT (distributed hash table); all routing is done entirely in userspace.

Breadcrumb routing

The router behind qaul.net is called “ratman” (route and transmission manager), which works via a gossip announcement protocol. Each node on a network periodically announces itself to it’s peers, letting them know over which of their local interfaces they can reach this node. Because of the coupling between ratman IDs and qaul.net users, this corresponds to announcing a user identity on the network for qaul.net, but different applications using ratman can bind these IDs to different semantics.

Let’s look at a small example:

  +--------> [ D ] <---------+
  v                          v
[ A ] <----> [ B ] <-----> [ C ]

Node A sends announcements to B an D, who will both proxy it to C. The router in C will use various metrics to decide which link is more stable, and declare it the “primary” for peer A. When C now wants to route a message to A, it looks up the local interface over which it thinks it can reach A the best (for example D). It then dispatches the message to D, hoping that this node will know where to deliver it.

No node in the network knows the full network topology, meaning that it can dynamically change, without greatly impacting routing performance.

There are several corner cases for what to do when a loop is encountered, or when a network becomes too big that it needs to be segmented. But this covers the basics of how ratman and qaul.net do routing.

Automatic roaming

Because qaul.net is primarily designed to run on unpriviledged mobile devices, a network automatically needs to be able to roam between transmission channels. In fact, because of technical limitations on mobile platforms connections between devices can’t be long lived and need to be broken up after a transmission, to sync with the next peer.

To make this handshake faster, ratman can sync the undelivered message buffers (called message journal) via a Merkel tree. This way mobile devices round-robin their connection states, each syncing undelivered packets to a neighbour, before moving on to the next peer.