Re: [v6ops] I-D Action:draft-ietf-v6ops-ipv6-cpe-router-bis-00.txt

[Mikael Abrahamsson <[e-mail]>]

On Fri, 18 Mar 2011, Brian E Carpenter wrote:

> Nothing, I think. Until we have clarity on the issues in 
> draft-v6ops-multihoming-without-nat66 I don't see what we can say, 
> beyond MUST support multiple prefixes and MAY support multiple WAN 
> interfaces.

A long time ago I did exactly this scenario, moved an enterprise from 
PA-subnet1 from ISP1 to PA-subnet2 at ISP2 over one month period, and both 
needed to work, and both ISPs did uRPF checking.

I normally say that whatever process came up with the answer "policy 
routing" (routing based on source IP address) you made a serious mistake 
way before you came to that answer, but in this case I think it's actually 
a valid mechanism.

In the draft I didn't see any mention of it, was it considered and 
rejected?

If one CPE has two WAN links and it gets two different PD prefixes, then 
it can send traffic to each one based on source IP address. How this would 
be discovered in a two WAN CPE router scenario I don't have an answer for. 
Whatever is done, it would also cause suboptimal routing for one of the 
prefixes since traffic would be routed via whatever default gw the host 
chose.

The major problem I see is how the two different CPEs would discover each 
others PD prefixes so they can do source based routing.

... or am I totally out in my thinking, perhaps this has been discussed 
and rejected before?

-- 
Mikael Abrahamsson    email: [e-mail]