1 · WelcomeRoot cause, instrument, prove — and where FrogNet came from.
The habits first, then the origin arc. Memory, not messages; the assembler-to-C argument; a live monitor across four machines. The 850 Kbps call is stated and immediately undercut: it is high, and you will do it at 450 and maybe lower.
2 · What it isA standalone, autonomous mesh — every FrogNet identical.
A 254-address segment per node, no permanent external control plane, and the sentence that sets the epistemic posture of the whole project: not the right way, not the only way — a first implementation. Then the Guild.
3 · HardwareWhat must be hardware, and what only must be software.
Debian. A Pi 4 with 2 GB is enough. An AP-mode router is required, and the one real restriction is stated plainly: we provide DHCP for the projected network.
4 · Installation, part 1frognet_install, the prompts, and the pond.
Recording scope: Recorded installer and broker setup. Current .run packages update an existing FrogNet node into region mode; use the current operator path for that deployment.
frognet_install, the --preserve flag, the broker prompts, setup_lillypad, reboot.
At the broker prompts: you will need an Internet-hosted machine to use the broker and enable cross-Internet network routing. If you have such a machine already, enter its address or hostname and a random port number (remember it) when requested during the installation process. A broker is always http://<host>:<port> — choose http, and leave the URL path empty. See the broker installation video for instructions on setting up a broker machine.
5 · Installation, part 2First boot: routes appear, and the node is on the network.
Recording scope: Recorded local/broker-era merge. Region-event application now uses shared facts, world and the graph; runMerge remains for boot/interface work.
run_merge, routes appearing live, a ping to New York, the node back online, and the same sensor record read from two machines.
At the New York ping: the machine we just installed was already WiFi connected to the Seattle5 site, which is a member of the author's personal network. This network already has a broker configured, so machines in New York City are available via Internet tunnel. See the broker installation video for instructions on setting up and configuring a broker.
6 · Broker installThe broker joins two independent FrogNets.
Recording scope: Recorded broker path. Current region membership and relay use the RAM manager and fdf-hub.
The bundle and its build script, frognet-pond create/list/status, and /etc/frognet/tunnel.conf. A rendezvous, not an authority — and a dependency at join time, which is said out loud.
7 · The Monitor · optional, strongly recommendedWatching the pond think.
Each machine writes its own state; the monitor just reads it — the same record from two viewpoints, and a live degraded peer diagnosed on camera. Its provenance is SCCM and OSTC.
This video is not required to get a FrogNet running. It is strongly recommended anyway: it is the best fundamental introduction to working with FrogNet Memory. The Monitor is not a management console bolted on the side — it is an ordinary reader of the same memory every program reads, and watching it is the clearest way to see what the programming surface actually gives you.
The Monitor has its own page →
8 · The CommunicatorThe ladder, on camera.
Recording scope: Recorded earlier Communicator and role machinery. The current desktop/Android implementation and per-viewer media path are described in the current chapters.
Bundles; the media, board-game and database hosts set at each merge; then the ladder demo — 1080 refused, 720 held, 854×480, 640×360, 480×360, 320×240, 160×120, and the climb back. Roughly four seconds down, fifteen up: the asymmetry is deliberate.