The loopback interface is simply a way of giving the router an ip address which is independant of the status of any physical interface. The advantage being that it is always accessable.
An example of its use would be a router running DLSW to 2 locally attached token rings. DLSW requires a local ip address and a remote ip address in order to tunnel traffic across a WAN. If the address on one of the token rings was chosen as the local address and that ring went down for some reason, then the other ring would be unable to send traffic across the DLSW tunnel, since the local ip address is no longer available. If a loopback address had been used, it would.
The other common use of loopback addresses is when running protocols like OSPF, which require the router to chose an IP address as the router id. Typically this will default to the highest ip address configured in the router and, again if this is on a physical interface and it goes down, this will caus problems. If however, the a loopback address is coded, it will be used instead.
The only disadvantage of using loopback addresses is that, like physical interfaces, they require a unique network address. This can be quite an overhead with classfull routing protocols like IGRP and RIP. With EIGRP and OSPF however, you can use VLSM and use a 255.255.255.255 address for all your loopback interfaces.
