RIP

Active gateways advertize their routes to others by broadcasting periodically a RIP message, which contains pairs of (IP address, number of hops), i.e. the IP address of a network and the number of hops along the path to that network. RIP uses a hop count metric to measure the distance to a destination. A gateway is defined one hop from directly connected network, two hops from a network that is reachable through an other gateway, three hops from a network that is reachable through two other gateways etc.

Passive hosts and other gateways listen and update their tables according to the broadcasted RIP messages. The routing tables will be updated with the smallest hop count to a particular network.

RIP updates may cause inconsistencies. Consider the following example :

The three gateways each have a route to network N1 (a). G1 is directly connected, G2 has learned it from G1 and G3 has learned from G3 from the RIP broadcast messages. When the connection between network N1 and gateway G1 disappears (b), G1 will set the hop count for the network in its routing table to infinity and report the new route in the next RIP message. But if G2 can broadcast its old value for hop count to network N1 (which is 2), just before G1 could send out the new RIP message with hop count infinity, G1 will update its routing table with a route through gateway G2 (i.e. will set the hopcount to N1 to 3), since the hop count of 2 indicates a shorter route than the direct route with hop count than infinity. Possible solutions to minimize the occurrence of a routing loop are:

• split horizon - a gateway should not propagate route information about a particular interface back to that same interface. In the example above, gateway G2 should not advertize its hop count 2 to network N1 to gateway G1.
• In case a gateway advertizes a short route to a network, other gateways update their tables immediately. E.g. if G1 is first connected to N1 it will broadcast a hop count 1 to N1. G2 and G3 will immediately update their routing tables using this information. On the other hand, if a gateway stops advertizing a route, other gateways waits some time, then find (if possible) an alternative route and broadcast that alternative. It could be that the alternative depends on the network that has gone down. However, at least the bad news travels slow, while the good news travels quickly . In the above example there is no alternative. So, after waiting some time G2 will broadcast to G3 the hop count infinity to N1.
• hold down - if a message is received that a particluar network is down, a gateway ignores new routing information about that network for some time. This gives other gateways time to receive the bad news and not mistakenly accept an out of date update message.
• poison reverse updates - once a network goes down, the advertizing gateway retains the entry for several update periodes, with hop count infinity. In other words, for some time it is specifically advertized that network N1 is down, rather than that network N1 is no longer included in the RIP updates to indicate that it is down. Poison reverse updates work best in combination with
• triggered updates -these updates force a gateway to immediately broadcast the recieved update, in stead of waiting for the next periodic update.

RIP uses UDP as the transport protocol and operates at UDP port 520.

The use of broadcast, potential for routing loops, hold down can make RIP inefficient for Wide Area Networks. For WAN other protocols are used for updating routing tables in the gateways.