[Autoconf] Another addressing model for AUTOCONF

[Alexandru Petrescu <[e-mail]>]

Addressing model we use, pdf 300Kb:

                  http://dl.free.fr/m95j1Km7a
(the username is left empty and password is 'password', without
  quotes.  File stays there for 30 days.)

Teco asked whether my draft contains an addressing model... true - it 
doesn't show so obviously
(http://tools.ietf.org/html/draft-petrescu-autoconf-ra-based-routing-00).

I said that there is an addressing model in this figure of the draft:
>                            egress|              |egress
>              ----     ----    ----              ----     ----    ----
>             | LFN|   |LFN |  | MR |            | MR |   |LFN |  |LFN |
>              ----     ----    ----              ----     ----    ----
>                |        | ingress|              |ingress   |      |
>               ---------------------             ---------------------
>                    2001:1::/24                       2001:2::/24

ThomasC and Chris also expressed doubts with respect to LFN--MR--MR--LFN
topology and link-local addresses; let me explain further.

We are using this addressing model on several moving networks. See the 
pdf at the beginning of this email. They show MR-to-MR with a single 
addressing scheme, then with a double addressing scheme; (double is 
necessary for our plan.)

And then a slide shows MR-to-MR-to-MR addressing model.

There are some scalability remarks and a route propagation model (pencil
and paper).

The mechanism has been prototyped and demoed since about one year now,
on three Mobile Routers and a bunch of LFNs, which shows it may work. We
have great plans for demoing on vehicles.

This is an addressing model we consider strongly. It needs later to
auto-configure some prefixes, because currently MNPs are pre-configured
in each moving network (this is the case in some deployments).

This addressing model is important to us, and uses link-local addresses.

Alex