Fault-tolerant broadcast of routing information

An algorithm is presented for the reliable broadcast of routing information throughout a network. The algorithm anticipates the possibility of long-delayed packets, line and node outages, network partitions, hardware failures, and a history of arbitrarily corrupted databases throughout the network. After any failure, the algorithm stabilizes in reasonable time without human intervention, once any malfunctioning equipment is repaired or disconnected. The algorithm also has the advantages of not requiring frequent control traffic in the absence of topological changes, not imposing artificial delays on nodes upon startup, and not relying on timers in ordinary operation. The algorithm is compared to a functionally similar algorithm in the ARPANET.