[Netext] PMIPv6 localized routing - IPv4 aspects

[sunseawq at huawei.com (Qin Wu)]

> Hi Qin,
> 
> please see inline for my thoughts.
> 
> Qin Wu schrieb:
>>> from previous mails and drafts, we can identify a couple of issues which we
>>> must take into account when designing a localized routing protocol for 
>>> PMIPv6.
>>> For the PS, I don't think we need to analyze problems which have been 
>>> addressed
>>> for standard RFC5213 operations already, such as a NAT between MAG and LMA.
>>>     
>>
>> [Qin]: Probably you are right. I think there are some misunderstandings.
>> Let me clarify why NAT between MAG and LMA is shown in the figure 1 of draft-wu-netext-pmipv6-ipv4-ro-ps.
>> I think NAT between MAG and LMA can be further divided into two case:
>> case 1: The MAG is behind the NAT
>> case 2: The LMA is behind the NAT
>> As regarding the Case 1, suppose there is one scenario where the MN attaches to MAG1 and CN attaches to MAG2, MAG1 and MAG2 register
>> to the different LMA, In this scenario, if there is NAT between MAG1 and LMA1 and the MAG1 is behind the NAT, I think also there is a NAT between 
>> MAG1 and MAG2, the MAG1 is behind the same NAT as the NAT located between MAG1 and LMA1.
>>   
> Yes, but that can be mapped to the case where there is a NAT between 
> MAGs, right? So it does not
> represent an extra case which introduces additional problems, or?

[Qin]: What I want to emphasize here is the scenario where the NAT is located between the MAG and the LMA is relevant to the case 1 and case 2 mentioned above if we consider local routing path setup between the MAGs or between LMAs.  In order to avoid confusion, I agree to  leave out this scenario where the NAT between the MAG and LMA from the IPv4 RO PS draft. Actually it does not reflected anywhere in the draft except in the figure.

>> As regarding Case 2,  as described in draft-ietf-netlmm-pmip6-ipv4-support-12, it is impossible for the LMA to locate behind a NAT unless the MAG and 
>> the LMA both locate behind NAT. Take the same scenario mentioned in the case 1 as example, if there is NAT between MAG1 and LMA1 and the LMA1
>> and MAG1 are both behind the NAT, I  think also there is a NAT between LMA1 and LMA2.
>>   
> If the IPv4 draft indicates that the LMA must not be behind a NAT, the 
> same holds for the
> RO protocol extension. Or is there anything extra to be analyzed for RO?

[Qin]: I agree. Does it mean RO PS draft needs one strong reason if it include the scenario where NAT is located between MN and CN's LMA?

>>   
>>> We could start with the following list to discuss and identify a 
>>> possible problem space
>>> for PMIPv6 localized routing and to find out which of these or new 
>>> issues are relevant
>>> for the PS and the NetExt protocol solution. I tried to collect 
>>> individual items from
>>> mail exchange with Sangjin, from draft-jeong-netlmm-pmipv6-roreq-01 and
>>> draft-wu-netext-pmipv6-ipv4-ro-ps-00.
>>>     
>>
>> [Qin]: As regarding IPv4 aspects, I am glad the draft-wu-netext-pmipv6-ipv4-ro-ps is credited. 
>>
>>   
>>> Any thought, comment and discussion is welcome.
>>>
>>> marco
>>>
>>> ----
>>>
>>> [1] MN and CN use IPv4 HoA for communication (IPv4 HoA mobility)
>>>
>>> [2] MN's and CN's MAG support different IP versions to signal to the LMA(s)
>>>
>>> [3] NAT between MN's and CN's MAG
>>>
>>> [4] NAT between MN's and CN's LMA
>>>
>>> [5] Different IP version for signaling between MN's and CN's MAG
>>>
>>> [6] Different IP version for forwarding of localized traffic between 
>>> MN's and CN's MAG
>>>
>>> [7] IP address conflict when MN and CN use the same IPv4 HoA
>>> Isn't that an issue with base PMIPv6 already?
>>>
>>> [8] Switch of forwarding tunnel from IPv6 to IPv4 when changing to
>>> localized forwarding between MAGs
>>> Should work if both MAGs are dual stack and can negotiate the IP version 
>>> (?).
>>>
>>> [9] Compatibility of route optimization states with IPv4
>>>     
>>
>> [Qin]: Thank for your summarizing and re-catogorizing the scenarios described in the draft-wu-netext-pmipv6-ipv4-ro-ps.
>> I think most of these scenarios you enumerated are reflected in the draft-wu-netext-pmipv6-ipv4-ro-ps except the case where the NAT is located between
>> MN's and CN's LMA. If my understanding is  correct, this case should require both MAG and LMA behind the same NAT. Because as described in the draft->> ietf-netlmm-pmip6-ipv4-support-12:
>> "
>> 4.  IPv4 Transport Support
>> the local mobility anchor must not be behind a NAT and must be using a globally routable IPv4 address. 
>> "
>> I wonder whether we need to address this case in the PMIPv6 localized routing PS draft.
>>   
> Probably not. In particular since NetExt scope is limited to RO within a 
> single PMIPv6 domain, where in
> a multi-LMA scenario these LMAs should be behind the same NAT. But as 
> you said, LMAs behind a
> NAT are ruled out, if my understanding is correct. 

[Qin] Actually I did not exclude the LMA behind a NAT if we allow both LMA and MAG are located behind the same NAT or there is any other concrete scenario.
But I wonder in muti-LMA scenario , why all these LMA should be behind the same NAT?
Is it possible for one domain to have several private networks or VPNs and each private networks has one LMA?

>But maybe there are 
> other opinions
[Qin]:  I look forward to seeing that if possbible.

> marco
> 
>>>
>>> _______________________________________________
>>> NetExt mailing list
>>> NetExt at mail.mobileip.jp
>>> http://www.mobileip.jp/mailman/listinfo/netext
>>>     
> 
>