Re: [MEXT] using MR-HA tunnel vs. combining BU // Re: WGLC for draft-ietf-mext-nemo-pd-01.txt

From: Romain KUNTZ <kuntz at lsiit.u-strasbg.fr>
To: Behcet Sarikaya <sarikaya at ieee.org>
Cc: mext <mext at ietf.org>
Sent: Monday, December 15, 2008 3:45:58 AM
Subject: Re: [MEXT] using MR-HA tunnel vs. combining BU // Re: WGLC for draft-ietf-mext-nemo-pd-01.txt

Hi Behcet,

On 2008/12/12, at 19:34, Behcet Sarikaya wrote:

We have an implementation of NEMO+DHCPv6PD, and here the "mechanism to determine the Mobile Network Prefix(es) owned by the Mobile Router" is a simple message from the DHCPv6 server to the HA that advertises the MNP/HoA pair to the HA.

[behcet] I am curious what is this message? Can you please give more details?

The DHCPv6 server is co-located with the HA on the same computer. I use a UNIX socket for inter-processus communication with a simple TLV message.

In our implementation, at home registration the HA checks with the DHCP server if a prefix is available for this HoA. If so, registration is done (but not forwarding). Once the DHCPv6 server has delegated a prefix to the MR, it informs the HA that it can setup forwarding for the delegated MNP.

[behcet] I have a question: If HA were a requesting router (not a delegating router as in Nemo-pd draft), how would you change your implementation architecture?
I have a feeling that you are using HA as a requesting router .

In our testbed, the HA is a delegating router, and the client & relay are located at the HA. DHCP messages are exchanged through the MR-HA tunnel.

If the HA were a requesting router, I guess the testbed, implementation and interaction with the mobility daemon would be completely different, so I can't really tell how I would change the current one.

If HA is a requesting router, all you need at the HA is DHCP Client (and possibly colocated with relay if the delegating router is not on-link)

I guess you refer to the solution exposed in draft-sarikaya-mext-prefix-delegation? This one suppose that the prefix delegation at the MR happens with BU/BAck messages, which is not the case in our implementation.

Note that I do not discuss here which solution is the best, but only present feedback from implementation of one of the solution. If I had chosen the delegation through the BU/BAck, I would certainly have coupled it with a DHCP server as you describe in your draft.

[behcet] and that would be much simpler implementation on HA simply requiring DHCP Client with PD capability. Such an implementation has other applications as well such as PMIP, so it could be reused. MR would be much simpler too.

Regards,

Behcet