Re: [DNA] New I-D: draft-vogt-dna-relocation-00.txt

[Christian Vogt <chvogt@tm.uka.de>]

> Later (step 5) the MN sends the BU from CoA to CoA2, which makes it
> do a NS/NA exchange to find the link layer address for CoA2.


Hi Erik,

the BU is sent to HoA2, not to CoA2 (to make binding updates robust to a
situation in which both peers move at the same time).

But I still think we are ok:  The first RO'ed datagram will trigger the
NS/NA exchange.  The NS carries the sender's MAC address in the TSLLAO
option, depending on whether it is sent by MN1 or MN2, respectively.

In summary, it doesn't matter if the nodes use Mobile IPv6 or not.  The 
mechanism seems to work also when the first messages exchanged between 
the on-link peers is a non-MIPv6 message.

- Christian

-- 
Christian Vogt, Institute of Telematics, University of Karlsruhe
www.tm.uka.de/~chvogt/pubkey/



Erik Nordmark wrote:
> Brett Pentland wrote:
> 
>> Ah yes.  For some reason a forgot that the MN would be the one 
>> initiating the communication.  Is there any way that an on-link 
>> node could obtain the MN's CoA and try to initiate a communication?
>> 
> 
> 
> I decided to walk through the details for the case when a CN on a
> link and the MN moves to that link, and does RO with the CN.
> 
> The steps would be: 1. MN uses DNA to detect that it has moved; sees
> a link up notification, and sends an RS (address is optimistic) 2.
> Receives an RA which indicates that it has moved. 3. Sends a BU to
> the HA. Initiates the BU process with the CN by sending a CoTI (and
> HoTI). The CoTI is sent from the CoA to the CN's address, thus it
> triggers the NS/NA exchange with the CN. 4. CN receives the HoTI and
> responds with a HoT. Receives the CoTI at about the same time. Tries
> to send a CoT. Has the neighbor cache entry for the CoA from the NS
> it received in #3.
> 
> Looks ok.
> 
> But if the CN was really a MN (MN2) (i.e., we have two MNs which have
>  moved to the same link), then the CoTI wouldn't be sent to the HoA2,
> and tunneled by HA2 to CoA2. Thus in #3 there wouldn't be a NS/NA
> exchange - the CoTI would be sent to HA2 (which we assume is off
> link).
> 
> But even in that case I think we are ok; the CoT will be sourced by
> MN2 with HoA2 as the source address, thus MN2 will reverse tunnel it
> via HA2. Hence MN2 doesn't need a neighbor cache entry in #4.
> 
> Later (step 5) the MN sends the BU from CoA to CoA2, which makes it
> do a NS/NA exchange to find the link layer address for CoA2.
> 
> I guess one can do similar analysis for other RO schemes, but it
> might be a bit tedious.
> 
> One can instead argue that if the MN sends the MLD report as part of
> #3, even if it is sent after the other messages in #3, then the
> probability that something will try to do a NS/NA exchange with the
> MN before the MLD report is sent is very small. And should that ever
> happen, then the fact that the NS will be retransmitted after 1
> second should take care of things.
> 
> So in summary, it's ok as long as the MLD report is sent "early",
> even if it isn't sent first.
> 
> Erik