Indirect Routing

When the network prefixes of the two addresses do not coincide, the destination
machine is in a different network. In this case, routing is carried out
(i.e., the most suitable gateway to direct the datagram to is chosen). This
routing is done by the host and the intermediate gateways. When a gateway
receives an encapsulated datagram through the physical network, it must
examine the IP address of the destination and decide which gateway of another physical network to send it to, carrying out the suitable encapsulation
for the new network.
To carry out indirect routing, the IP specifies an algorithm based on
tables. This is discussed next.
3.5.2.1 The IP Routing Table Algorithm
The IP specifies a routing algorithm based on tables stored both in hosts and
in gateways. An example of a routing table is shown in Figure 3.7.
It is obvious that a gateway only knows the next step toward the destination
network and not the complete path. Bearing this in mind, and knowing
that only the network prefix of each destination IP address is stored (the
suffix specifying only the machine used during the final direct routing), the
routing tables can be maintained with a reduced size. It should also be
remembered that in this table a gateway or host only stores the gateways connected
to its own physical network, which is always the first step toward any
other destination network.
This routing mechanism has some drawbacks. For example, all of the
datagrams directed to a specific network will follow the same path even
though other alternative routes could be used at the same time. Moreover,
the influence of the delays and the network throughput are not taken into
account according to the type of traffic.
What happens when a destination network does not appear in the
table? In this case, there is what is called a default gateway, which routes all of
the datagrams whose destination network does not appear in the table. A
default gateway is often the only entry in the routing tables of the hosts.
Finally, another important aspect of routing is how the tables are initialized
and how they are updated as the network changes. In the beginnings
of Internet, these tables were updated manually in the few gateways that
existed, but this situation became impracticable. A series of protocols then
developed as the Internet grew and changed its architecture permitting the
initialization and updating of these tables automatically and dynamically
through the interchange of routing information among the gateways.