Hi Satoru,

In addition to my previous email (not the one you last replyed): DHCPv4
over IPv6 will create/refresh the per-subscriber state. And LW4over6
features to keep it dynamic(on demand).

Thanks,
Qi

On Wed, Jun 27, 2012 at 7:42 PM, Qi Sun <[email protected]> wrote:

> Hi Satoru,
>
> Please see inline :)
>
> Qi Sun
>
> On Wed, Jun 27, 2012 at 6:27 PM, Satoru Matsushima <
> [email protected]> wrote:
>
>> Qi,
>>
>> On 2012/06/27, at 19:01, Qi Sun wrote:
>>
>> >
>> > [Qi] DHCPv4 over IPv6 is a provisioning method. And it's about the
>> public IPv4 address allocation, NOT about IPv4 address and IPv6 address
>> mapping. So there is no state. Please read the draft of DHCPv4 over IPv6
>> for clarification.
>> > LW4over6 needs to maintain the binding table on TC. That's stateful.
>>
>> That sense is 'stateless' in my mind if there's no DHCP state in the BR.
>>
>> [Qi] What discussed here is about states on data plane, aka the v4-v6
> addressing binding(please refer to joel's email). The binding table could
> be dynamic(on-demand) or static.
>
>
>> > Actually in MAP 1:1 mode there will be a binding table right? Here
>> comes the question: what will you do if there are new comers or some users
>> leaving the network?
>>
>> it is just an operation in provision. Now I find that LW46 could be
>> possible to merged with MAP, or vice versa. thanks.
>>
>
>
> [Qi] LW4over6 is a solution targeting at the per-user stateful/binding 
> scenario, pure and clear.
> MAP is originally targeting at pure stateless/alogrithmic address mapping 
> scenario, and then comes
> to this 1:1 mode due to the theoretical corner case of EA-bits=0, according 
> to you guys.
> While I believe the main targets are differnt and paralled
> (actually two out of three in 4-over-6 case,
> with the 3rd one per-session stateful), I don't think merging is the proper 
> way to proceed.
>
> Thanks,
> Qi
>
>>
>> cheers,
>> --satoru
>
>
>
_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires

Reply via email to