Re: radical solutions

[Masataka Ohta <mohta@necom830.hpcl.titech.ac.jp>]

Sean;

> | But, why do you insist that we have to change routing for
> | multihoming?
> 
> We need to change routing anyway to support more complex expressions
> of policy that allow network operators to constrain certain traffic to
> certain subsets of their local topology.

You are a simple guy and do not handle complex expressions of policy.

That's the way to keep the Internet operational.

> This is driven in part by the 
> sort of greedy commercial interests that employ people like me. :-) :-) :-)

You don't have to do anything technically and should simply tell the
commercial people that the issue is so complex that they should pay
more salary for you forever.

> | End-to-end multihoming does not affect routing.
> 
> Unfortunately, I did not fully understand your draft which explains it.
> I am a simple guy, and might benefit from a simpler summary.

Let routing concentrate to compute singly homed routes.

Give applications all the addresses and let applications try all of
them based on the application specific time-out periods.

Local routing table and ICMP may or may not help to skip hopeless
addresses.

It is already done for DNS and SMTP servers.

						Masataka Ohta