Re: [nemo] WG Last Call on draft-ietf-nemo-terminology

[Romain KUNTZ <[e-mail]>]

Hello Chan-Wah,

Chan-Wah Ng wrote:
>>A split NEMO occurs when a single mobile subnet connected to multiple 
>>Mobile Routers, or multiple mobile subnets interconnected with routers, 
>>become two or more independant mobile subnets.
> 
> 
> "independent".
> 
> Anyway, "multiple mobile subnets interconnected with routers" is not
> very accurate and leads to controversy.  Consider
> 
>     MR1            MR1
>      |              |
>   ---+-+---  =>   --+---   MR2
>        |                    |
>       MR2                 --+--
>        |
>     ---+----
> 
> It fits your definition but does not falls into the normal sense of
> split NEMO.

That's right.


>>     |             |          |       |
>>    MR1  Router   MR2        MR1     MR2
>>     |    |  |     |    ->    |       |
>>   --------  -------        -----   -----
>>
> 
> 
> With a router, presumably a LFR, in between, the two links are separate
> subnets.   So, MR1 and MR2 are handling different prefixes.  When the
> two links goes separate way, what would happens is that the LFR would no
> longer function as a router, since it loses its at least one of its link
> for forwarding.
> 
> As far as NEMO-BS is concerned, this situation also does not fall into
> the normal sense of split-NEMO problem, because MR1 and MR2 are handling
> different prefixes, even when they split, there is no ambiguity in the
> HA(s) on who to forward a packet destined for one of the MNPs to.

Correct. I do not know what I had in mind when drawing this :-)
I was thinking about MR2 registering some of MR1 prefixes for redundancy 
purposes, but this has no relations with current NEMO-BS.

> It all depends on what is the motivation for the inclusion of the
> split-NEMO definition.  I thought the motivation was to assist in
> illustrating the problem of split NEMO as documented in
> draft-ietf-nemo-multihoming-issues.  If this is indeed the case, then
> the definition of split-NEMO should define cases only when the problem
> of split NEMO occurs.

I agree. So do you see other case than (n,1,1) with the following topology:


     |    |             |      |
    MR1  MR2           MR1    MR2
     |    |      ->     |      |
   ----------         -----  -----

(ie multiple MRs attached to the same mobile subnet on their ingress 
interface, registering the same prefix to the same HA)

where split NEMO is a problem?

Romain