Hello Edison,
I would like to know your opinion on the "ip default-network". In my opinion, there are various counterintuitive facts about it and the entire command bears too much annoying cruft and dependencies from the ages of classful routing protocols like RIPv1 and IGRP for which it was originally designed. Its behavior is quite complicated, classful-address oriented and dependent on the contents of the routing table in the moment of its use.
Assume that you enter the command
ip default-network A.B.C.D
where A.B.C.D is an arbitrary IP address. Now, a couple of different scenarios exists.
1.) If the routing table does not contain a classful network (I will call it major network) of the A.B.C.D nor any of that major network's subnets when issuing that command, it will be added to the configuration exactly as entered. For example, if the routing table does not contain the network 1.0.0.0/8 nor any subnets thereof, the command
ip default-network 1.2.3.4
will be added to the configuration as shown here. Note that it references a network that the router does not know about (that was the premise in this case). However, as soon as the network 1.0.0.0/8 or any of its subnets appears in the routing table, the router will automatically remove this command from the running configuration and replace it with a placeholder static route as explained in the case 2.) - see below.
2.) If the routing table contains a major network of the A.B.C.D or any of that major network's subnets, the command will not be added to the configuration. Instead, a placeholder static routing entry will be created in the form
ip route MAJOR_NET(A.B.C.D) CLASSFUL_MASK(A.B.C.D) A.B.C.D
The command "ip default-network A.B.C.D" in this case will be accepted without creating that placeholder static route if and only if the A.B.C.D is itself a major network.
For example, if there is a network 1.1.1.0/24 in the routing table, issuing the command
ip default-network 1.1.1.0
will cause that a different command is added to the configuration:
ip route 1.0.0.0 255.0.0.0 1.1.1.0
The command "ip default-network" will be accepted only if written in the form:
ip default-network 1.0.0.0
That static "ip route" command is created because of limitations of original IGRP. The IGRP does not support the default route in the form of 0.0.0.0/0. It can only flag a specific network as a candidate for default route. The creation of that placeholder static route provided a network entry in the routing table that had to be redistributed into IGRP (note that it is classful) and only then flagged as candidate for the default route.
3.) Any supernets that would cover the IP address A.B.C.D are ignored and the command "ip default-network A.B.C.D" behaves identically as in the case 1.)
So the "ip default-network" is a heavily classful-oriented command which makes it kind of difficult to use in today's intensively subnetted networks. Moreover, OSPF and IS-IS ignore this command. From my viewpoint, it was only required for IGRP that had no other means of conveying a default route. For all other routing protocols, there are other (and in my opinion, better) means to inject a default route. Its operation is too cumbersome and I have the feeling that it is a remainder of a past that is more confusing than helpful.
Best regards,
Peter
