Re: [v6ops] I-D Action: draft-townsley-troan-ipv6-ce-transitioning-00.txt

[Brian E Carpenter <[e-mail]>]

On 2011-12-18 16:59, Fred Baker wrote:
> On Dec 13, 2011, at 5:21 PM, Brian E Carpenter wrote:
> 
>> It seems to assume that
>> ingress filtering is inevitable, but this is very restrictive as far
>> as seamless multihoming goes. If we can one day get to a situation
>> where ingress filtering is automatically tailored for multihomed customers,
>> we will need each SRIB entry to point to multiple DRIBs accordingly.
> 
> To be really honest, I think you need to demonstrate that there is interest in achieving the nirvana you desire. Generally speaking, a service provider will do what you pay him to do, and anything that he can't describe in generic "service" terms will be generally offered, if at all, at a high price. Couple that with the general behavior of markets - seeking the lowest price for a reasonable set of services, and...

Well yes, the market does what it does. If small users come to care
deeply about multihoming, the incentive for ISPs to support flexible
ingress filtering will arise and a solution will emerge. If the users
don't care, there will be no incentive.

My point is that we shouldn't write specs that unnecessarily constrain
things that might be desirable. I think a tiny change will do that:

   MH-2:  An IPv6 CE router MUST create an SRIB containing entries for
          associated delegated prefixes.  Each entry points to one or
          more DRIBs.  An entry points to multiple DRIBs only in the
          case where an identical delegated prefix is known to be
          routeable via multiple WAN interfaces.

 Brian


   Brian