Best afterPacket switching

Fault tolerance

The internet keeps working when pieces of it fail because there is usually more than one path between any two points. Fail a node on a redundant network and the traffic reroutes; fail the one node a single-path network depends on and everything past it goes dark. Switch the abstraction on to watch a node die (the ✕) and the packet find — or fail to find — a way around.

Send a packet from A to D after node B fails. A redundant network should find another way.

Fault tolerance — the single point of failure execution-derived · CPython model: author-asserted

9            at = nxt10            path.append(at)11            moved = True12            break

same data, two zoom levels

state after line 1 runs

path: —

step 1 / 11

Click a line, drag the slider, or use the ← → keys.

What you are looking at

In both versions node B has failed, and the router walks the graph picking the first live neighbour toward D. The only difference is the shape of the network.

Buggy — a straight line A–B–D. A's only link is to B, so when B dies the packet is stranded at A and never reaches D. B is a single point of failure.
Procedural — a redundant network where A also connects to C. The router skips the dead B and reroutes A → C → D. Same failure, but the extra path saves the delivery — that redundancy is exactly what "fault tolerant" means.

The packet's path is real execution output; the little network is a hand-built model (marked model: author-asserted) — see how GlassBox stays honest.