[e2e] Loopback address and end2end implications

[Vinay Bannai]

I apologize in advance if some folks find this email not directly relevant
to end2end mailing list. 

Using loopback address seems to be common technique used to guard against
link failures and interfaces going down. However using loopback addresses
tends to bloat the routing tables as they are advertised as host routes
(atleast the ones that I am familiar with). On the other hand, if I have a
physical interface which does not go down inspite of LOS or link outage (it
is like a loopback but it is tied to a physical interface), does it make
sense to withdraw the subnet route and still keep the host route when there
is a link outage? This should make the interface reachable via other
interfaces on the router. The primary reason for doing something weird as is
this to reduce the number of loopback addresses advertised and keep the
routes to a smaller size. Only in the event of a problem with the physical
interface (LOS or link outage) it morphs into a loopback-like address. 

I looked up the archives of FreeBSD and other mailing lists but did not find
any reference to such a optimization. Does it have any implications end2end
that I have overlooked?

Thanks
Vinay