Re: [MEXT] DHCPv6 relay co-located with the DHCPv6 client at the MR for DHCPv6PD in NEMO
Alexandru Petrescu <alexandru.petrescu <at> gmail.com>
2010-05-05 16:42:46 GMT
Le 05/05/2010 18:31, Behcet Sarikaya a écrit :
> Hi Alex,
> Sorry for my late reply.
>> Hi Behcet and thanks for the pointer.
>> It is not clear to me whether the 2007 draft you cited
>> addresses the problems listed in the recent discussion.
>> That draft uses BU extensions (new BU bit) as part of the MIPv6
> protocol. I
>> disagree with it, because DHCPv6-PD is better than DHCPv6
> at allocating
>> prefixes, and already implemented, and already used. I do
> not see why
>> extending MIPv6 to allocate prefixes.
>> To me (IMHO) a good solution would use DHCPv6-PD non-encapsulated,
>> between MR and AR-Relay.
> Absolutely. This was missing in draft-ietf-nemo-prefix-delegation-02.
> That is the reason why we redesigned it using DHCPv6 PD, but using it
> from HA as RR to the delegating router as in
But this draft still extends BU/BAck to do PD on MR (it's in the title),
instead of using DHCPv6. I do not agree with this.
This draft does use DHCPv6-PD for HA as RR (DHCP Requesting Router, not
"Return Routability" nor "Resource Records" I agree with this in
general, although not sure whether HA could be RR if MR were too -
better have a single RR in the picture.
>> Generally speaking, if I had to describe the need:
>> -need to allocate dynamically an MNP to the MR.
> -need to perform it securely: some of the risks are: (1) an MR
> falsely request a prefix prior to having identified itself to HA
> its MIP6 keys, (2) the HA may miss the allocated MNP from its
> Table and thus couldn't ensure the MNP in the explicit mode BU
>> -need to reuse the NEMOv6 MIP6 protocol unmodified. -need to reuse
>> existing protocols naturally fit for prefix allocation.
>> -reliability: need to allocate a prefix to MR even when the MIPv6
>> implementation fails; this is good in order to allow in-mobile
> communication between LFNs; MIP6 may fail when the MR is
> and when MIP traffic blocked by firewalls in the access
> -need to inform the HA about the MNP allocated to the mobile
> such that NEMOv6 MIP6 implementation works unmodified, and is
>> able to
> check the presence of the prefix in the NEMOv6 Prefix
> I think we need a fresh look into this problem and what you have
> above could be a good starting point.
WEll we have a WG item, suffices it to change it according to remarks on
the WG list. Or according to its consensus which should yet to be found.
> It is a difficult problem.