Re: (v6ops) WG Last Call: draft-ietf-v6ops-renumbering-procedure-00.txt (fwd)

[Fred Baker <fred@cisco.com>]

At 12:00 PM 07/08/04 +0200, Brian E Carpenter wrote:
> While you are tweaking, it might be worth adding a note that the eventual
> multi6 solution(s) may add new issues to the renumbering recipe, or
> alternatively may simplify it.

Actually, I'm not at all sure that they can.

The problems with renumbering without a flag day are not in the technology
there to assign new addresses and such. They are first and perhaps foremost
in the applications and configuration scripts that forego those tools in
favor of static or configuration.

More subtly, there are issues with the "who knows what when" operational
steps that ensure that a one system knows a given address of another system
if and only if the other system is using it and the connecting
infrastructure in fact connects them. If you have selected a new address
but one of the eleven routers between you and I doesn't know how to route
to the prefix, when I use your new address I experience a disruption. If an
address is being advertised for you, or was advertised yesterday with a TTL
that leaves me with an active record today, and you stop using the address,
I experience a disruption. I'm talking about making this work not only in
theory in a corner of the world, but operationally supporting applications
using them on an end to end basis.

But as you ask, I will drop the Multi6 copy after this note.