UNIX System Administration Handbook - Evi Nemeth [217]
The version statement tells whether to run RIP-1 or RIP-2 on the specified interfaces. The default when running RIP-2 is to multicast updates rather than broadcast them, which prevents RIP-1 routers from seeing them. You can specify broadcast to force broadcasting to occur.
By default, gated will listen to RIP updates from anyone who sends them. If a list of trustedgateways is present, however, gated will only pay attention to the listed hosts. A gateway_list is just a series of IP addresses separated by whitespace.
sourcegateways are hosts to which RIP updates should be sent directly, rather than via broadcasting. This feature can be used to reach hosts on different networks or to target routers on a network that does not support broadcasting (or on which broadcasting has been disabled).
traceoptions are specified as described on page 355. Any options included here will apply only to RIP. The RIP-specific packet-tracing options request, response, and packets log requests received, outgoing responses, and all packets, respectively. Packets are normally summarized. If detail is specified, a more detailed dump of each packet is included in the log.
An actual RIP configuration clause appears in section 1 of the complete gated.conf config file example on page 365.
Some preliminary background on OSPF
Before launching into the gory details of OSPF configuration (which are in truth not really so gory), we need to talk a bit about two more features of OSPF: routing areas and designated routers.
Routing areas
At a large site, it may not be necessary or desirable to distribute a complete set of link states from one corner of the network to another. To cut down on the amount of update traffic, OSPF allows individual networks to be grouped into “areas.” Link-state information (i.e., information about the network’s physical topology) is propagated only within an area; information about the area is distributed to the outside world in the form of routing summaries.
Every network is a member of exactly one area, and areas can include more than one network. Routers are considered to be members of all the areas on which they have network interfaces. A router that belongs to more than one area is called an area-border router and is responsible for translating link-state records into summary records.
A routing summary is really just a collection of routes: “Router X can send packets to network Y in 3 hops,” where X is an area-border router. The routers outside an area combine the declared summary cost with the computed cost to the area-border router to determine a total path cost to the network.
This scheme might sound like a distance-vector routing protocol in disguise, but there are two important differences. First, summaries are propagated exactly as they came from the area-border router that originated them. A router may compute that if it is two hops away from X and X is 3 hops from Y, then it must be 5 hops away from Y. However, it will never reveal the result of this calculation to another router. It will just pass along the original summary route.7
The second difference from a distance-vector protocol is that the OSPF scheme does not attempt to deal with arbitrary network topologies. OSPF requires that all routing areas be logically adjacent to a central backbone area known as area 0 (though they may be adjacent to each other as well). Route summaries can travel only from a leaf area to the backbone, and vice versa, not directly between leaf areas.8
This simple two-level hierarchy forestalls the possibility of loops.
If your real-world network architecture does not match OSPF’s two-layer model, all is not lost. You can still represent it as a two-layer hierarchy by using an OSPF concept called “virtual links.” Unfortunately, a discussion of virtual links is beyond the scope of this book.