A network you learn
by running it.
NATLAN is an independent network on DN42. It exists to explore routing by doing it: keeping real sessions up, looking after real machines, and connecting with other people who do the same.
Why it exists
NATLAN started in November 2025 as a couple of machines joined by wg-quick, configured by
hand over SSH with the keys kept in a notes app. That held for two nodes; the third one broke it. The
network has been rebuilt twice since — first on BIRD templates and deploy scripts, then on a control
plane with an agent on every node — and each rebuild taught more about routing than reading ever did.
The aim is a hobby network run like a real one: sessions that stay up, filters that mean something, changes that are written down. It grows a site at a time, and peering is open to anyone on DN42.
How it is built
- SITES Eight sites in eight cities. Frankfurt, Xi'an, Guangzhou, Shanghai, Hong Kong, Taipei, Tokyo and San Jose, each its own place on the map; six up right now.
- BACKBONE WireGuard between the sites, eleven backbone links in all, with iBGP carrying routes across every site.
- PEERING BGP with other DN42 networks, over WireGuard; routes that fail ROA validation are rejected at import.
- RECORD A journal of what changed. New sites, rebuilds and outages get a dated entry.
How we run it
Who runs it
One operator. NATLAN is registered in DN42 under the NGWORKS organisation (maintainer NGWORKS-MNT), and every change to it — a new peer, a rolling reboot, an agent release —
goes through the same control plane, so there is always a record of what changed and when.