Re: [Softwires] Proposed Unified Address Mapping for encapsulation and double-translation

[Jacni Qin <[e-mail]>]

hi Satoru, Remi,

On 10/11/2011 5:47 PM, Satoru Matsushima wrote:
> ...
>>> Since Remi's draft doesn't specify 'CE IPv6 prefix length', IPv6 prefix can't be self delimiting to extract IPv4 address and port-set ID in the case of double translation. That's my understanding. Is that correct?
>> In my understanding, yes.
>>
>> Now, putting the "CE IPv6 prefix length" back as a rule parameter shouldn't be a problem if the idea of linking IPv4 and IPv6 sizes of address spaces is abandoned (a choice that, I have to agree, simplifies the design).
Not "abandoned" actually, just not that tightly :-). L is not necessary 
by a CE when deriving its v4 space, since it would have got the length 
of prefix delegated to it (so, the L in the rule picked should match the 
length of delegated prefix). The sharing ratio is determined together by 
the L (Rule IPv6 prefix), L (Rule IPv4 prefix) and the L (delegated 
prefix). For example, for a given L (delegated prefix), the other two 
"L" can be changed to get different sharing ratios, or to maintain an 
identical sharing ratio within a domain even the L (delegated prefix) is 
variable.

L (in the matching rule based on IPv4 destination address) allows the CE 
to know the exact number of bits to pick from the port when forming the 
IPv6 destination address.

>>
>> Cheers,
>> RD
> I think that it should be exclusive in a domain but it could be coexist in the map specification.
Sorry, I don't quite follow, can you please say a little more?


Cheers,
Jacni

>
> cheers,
> --satoru
>
>