Re: [MEXT] Last Call: draft-ietf-mext-nemo-pd (DHCPv6 Prefix Delegation for NEMO) to Proposed Standard

[Alexandru Petrescu <[e-mail]>]

Le 10/09/2010 18:57, Laganier, Julien a écrit :
> Alexandru Petrescu wrote:
>>
>> Le 10/09/2010 11:58, Hesham Soliman a écrit :
>>>
>>> On 10/09/10 7:55 PM, "Alexandru
>>> Petrescu"<[e-mail]>  wrote:
>>>
>>>> Le 10/09/2010 11:48, Hesham Soliman a écrit :
>>>>>>>
>>>>>>> =>     Who cares, specify it in your product
>>>>>>> description. The IETF doesn't specify how to build
>>>>>>> products.
>>>>>>
>>>>>> Hmm... to me it is a very IETF sensitive issue the Router
>>>>>> vs Host. For example, an ND spec says distinctively what a
>>>>>> Host and what a Router does, e.g. a Host does not respond
>>>>>> to Router Solicitation.
>>>>>
>>>>> =>    Yes and it does so on a per-interface basis, not on a
>>>>> per-machine basis.
>>>>
>>>> Yes, and the Mobile Router is a Router on its egress interface
>>>>  when connected at home, as per spec.  It is that interface
>>>> that needs a default route automatically configured.
>>>
>>> =>   Ok, so you're happy with it being half host half router
>>> when it's away from home?
>>
>> When it is away from home it is fully a Host on the egress
>> interface. When at home fully Router on same.  I am happy with it
>> this way.
>>
>>> If so then let it do the same at home. Otherwise, I don't know
>>> how you want to fix this in this WG.
>>
>> It would mean to specify it to be a at home, be first a Host (get
>> default route) then change and become a Router, but still at home.
>>
>> This behaviour could be set in the DHCPv6-PD-NEMO draft, being
>> under discussion now.
>
> This is a non issue for this draft. This is not specific to NEMO but
> generic to any DHCPv6 Prefix Delegation setup where the requesting
> router needs to configures an address on its north side.

Right... and a default route - that's the blocking point to me: I have
all this shiny software and RFC but no automatic default route, I have
to manually configure it.

> It can do so as per the deployment specifics, including, but not
> limited to, acting as a host on the north interface and as a router
> on south interfaces -- please remember that Neighbor Discovery is
> specified on a per-interface basis.

It's not sufficient to say Mobile Router always acts as a Host on the
North interface.  Because it must act as a Router too on its North
interface.  It must join the all-routers multicast address on the home
link, in order to forward packets from hosts on the home links towards LFNs.

If you don't like my default route comment then you could also just
write in DHCPv6-PD-NEMO "this does not configure a default route on the
Mobile Router, neither SLAAC".

> [ I also note that this draft has been more than 2 years in the MEXT
> working group in which you are participating, which gave you ample
> time to comment on this and other things... ]

Note taken and I owe explanation.  If you wish consider my comments
non-blocking, as comments to a soon-to-be-Request-For-Comments.

I simply did not have neither the software nor the network testbed to
prototype until now.  Obtaining that is not easy, takes planning and
time. E.g. a stable DHCPv6-PD was available only lately from ISC despite
early initial availability.  It was mainly written for ISP-to-House
deployments which are very different than Mobile Router moving around.

If you wish too, you could bring it back from IESG to here and I will
comment further on it to work with real Relay, to solve
the sync between Prefix Table, Routing Table, radvd.conf, lease file, to
require Relay to insert route (currently it doesn't!), to respect 'M',
and so on.

Or we could just deliver to IESG and see past.  I am fine with all
options.

Alex

>
> --julien
>