[c-nsp] NSSA default route forwarding address and default route selection

[Justin M. Streiner]

On Tue, 5 Dec 2006, Dimitrios Kalogeras wrote:

> Hi to all of you,
>
> I have the following scenario:
>
> A router which is a member of the area 0 and member of NSSA  generates
> default route inside the NSSA. The  router has two links  and no
> loopback in  which belong in this NSSA area. The router selects as
> forwarding address the biggest IP address and not the one with the
> smallest ospf cost nor does he selects 0.0.0.0 as a forwarding address.

This is correct.  If an OSPF speaking router does not have a loopback 
address, it will use the highest-numbered IP address that it has on one of 
its interfaces as its router ID.  If the router in question has many 
interfaces or its overall topology changes often, you could introduce some 
IGP instability if a new interface is added with a higher IP address than 
the existing router ID.  This would cause your OSPF adjacencies to drop 
since hello packets would be sent out with a router ID different than the 
ones that originally established the adjacency.  It would also likely 
cause your OSPF process to reset.

Having a loopback address on your routers is a good idea for many reasons, 
but if for some reason this is impossible, your options are 1) live with 
the limitations described above, or 2) use the "router-id" command under 
"router ospf XXX" to tie the router ID to a specific interface.

> Can you explain this behavior ? Is this rfc specific or implementation
> specific ?

I believe this is part of the OSPF protocol.  It's been awhile since I 
read the spec, so things might have changed in the past few years :)

jms