Protocol in Code · Track
distance-vector rumor propagation and its count-to-infinity failure mode
Sessions
An OSPF Router-LSA is a fact a router asserts about its own links, flooded unchanged across the area. A RipRoute is neither of those — it is one neighbor's claim about hop count, already folded together with everything that neighbor believed before it. What does it mean to trust a number like that, and what does the code let you check before you do?
OSPF's SPF run in ospf/spf.py is Dijkstra over the whole LSDB — a graph object, a priority queue, one full recompute. RIP never sees the graph at all: it only sees what one neighbor just said. What does it mean to run Bellman-Ford when all you have is a single pass over a list of (prefix, metric) pairs, and what has to be true before a new number is allowed to overwrite an old one?
RIP needs a metric value that fits in the same integer field as every real distance but means "not reachable at all." Why is that value 16 — not 255, not a billion — and what does picking a small number actually buy?
Open Session 03Why does a distance-vector router that just lost a route end up believing its own rumor came back from a neighbor, and how many rounds does that belief take to burn itself out?
Open Session 04If the cure for count-to-infinity isn't smarter math, what is it, and where exactly does it run?
Open Session 05What does the smallest readable RIP speaker look like when a routing table, a neighbor list, an outbound filter, and a Bellman-Ford pass are all wired into one object that a round-robin exchange loop can drive to convergence — and how is that whole shape different from OSPF's?
Open Session 06