IPv6 multihoming question

[Steve Powell]

Greetings.  The multihome addressing issue for IPv6 is definitly
interesting and I sympasize with the idealists who would like
to see nice neat, clean routing tables.  Then again everything I
have always had an idea for started neat and clean.  Before I 
knew it, lifes one thousand exceptions had blown it up into
a monster, and they come shambling out of the garage to eat me. ;)

In my opinion the issue boils down to a statistical mechanics
problem.  If additional information is to be transmited(multihomed
customer), then additional "state" is required to support the
transmission of that information.  Regardless if you use the Ipv4
solution, the shim draft, DNS, or even build  the mechanisim in
the client/server, something must carry the additional information
and make decisions.  The superior solution will carry the most
information and generate the least amount of entropy.

The IPv4 world likes big messy routing tables because ultimately
this centralized model carries the most information for the least
amount of entropy, which is a byproduct of state.  Previous 
submiters have pointed out that router vendors are building more
capable devices that are able to easily handle projected route
table growth. 

The IPv4 world is very happy with how IPv4 works along with their
ethernet.  The IPv4 world will not convert to IPv6 unless the IPv4
world observes the same luxuries available in IPv6.  If we wish
to turn the world upside down, we have to show a compelling
reason for doing it.  Since big routing tables is an issue easily
handled by the IPv4 world, arguments for doing multihoming differently
are not going to sway it.

stevep