Re: delayed multihoming/mobility set-up

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

Pekka Savola;

> Me and Dave Crocker sat down and thought about multihoming and mobility a
> bit.  Among other things, we came up with one notion which may not have been
> explicitly mentioned, but may be relevant:

>   Is it necessary to ensure connection survivability prior to the connection
>   establishment?

Survivability of what?

Nonexisting connection?

If the attempt of establishment fails for some reason, there can
be no survivability.

Abortion is leagal, here.

>   Or, is it necessary to ensure connection survivability in parallel with
>   the connection establishment?

Because of race conditions, it does not ensure anything.

A conservative leagal request is to ensure the (best effort)
survivability at birth, which requires piggy backing and minimal
(3way) handshaking, which means there is no race.

>   Or even, is it OK to just ensure connection survivability only for
>   "long-lived" sessions (e.g., those which have lasted for longer than 5
>   minutes), using some definition.

There is no proper timeout value at the IP layer.

> The last one would be obviously useful with low-mobility rate, or site
> multihoming when the connection survivability method you're using would
> require extra packets or extra delay to set up -- and you'd want to avoid
> that when it's not necessary.

You are trying to import "connection" at the IP layer, in vain.

"low-mobility" is rather a link layer issue.

							Masataka Ohta