Re: [Softwires] ietf-softwire: IPv4 + PSID primary key for lw4over6 binding

[Andy Wingo <[e-mail]>]

Hi Ian,

Thanks for being so patient with a newcomer.  A doubt inline:

On Tue 12 Jul 2016 15:00, Ian Farrer <[e-mail]> writes:

>> On 12 Jul 2016, at 14:48, Andy Wingo <[e-mail]> wrote:
>> 
>> I am new to YANG; apologies in advance for making all of the beginner
>> mistakes.  My understanding of the specification
>> 
>>              list binding-entry {
>>                key "binding-ipv6info";
>>                description "binding entry";
>>                uses binding-entry;
>>              }
>> 
>> was that "binding-ipv6info" uniquely identifies the B4 (because it's a
>> key within the binding-entry list).  Is that not the case?  If it is the
>> case, how is it possible for one B4 to have multiple softwires?
>
> [if - It depends what defines multiple softwires here. The B4 can have
> multiple softwires if the source v6 and the v4/PSID tuple are
> unique. The routing choice for the B4 of which softwire to use for which
> traffic is local to the implementation and not specified.
>
> From the lwAFTR’s perspective, they are two separate binding table
> entries. The fact that they are going to the same B4 is invisible.]

I'm definitely on-board with the possibility of there being more than
one softwire per B4.  At least from the perspective of the AFTR that is
entirely compatible with RFC 7596.

However it still seems to me that the YANG module prohibits this from
being the case.  The only key in the "binding-entry" list is the
binding-ipv6info, which of course is just an IPv6 address or a prefix
identifying the B4.  Because the key must be unique within the list,
unless you have multiple distinct binding-ipv6info values for a B4 you
can't have multiple softwires per B4.

But, I could be reading the draft YANG module and/or RFC 6020 entirely
wrong.  I look forward to being less ignorant :)

Andy