Re: A common problem with SLAAC in "renumbering" scenarios

[Philip Homburg <[e-mail]>]

In your letter dated Mon, 4 Feb 2019 13:53:19 +0000 you wrote:
>Philip Homburg wrote on 04/02/2019 13:47:
>> In theory we could specify how access routers could do the right thing. And 
>then
>> get CPEs to do the right thing as well.
>> 
>> But I think it would be more productive to make hosts more robust.
>
>making hosts more robust is a good deal more difficult and error prone 
>than making ISP provisioning systems more aware of maintaining state.

It seems to me that setting the preference time to zero for all addresses
derived from a PIO from a particular router that are not covered by
the PIOs in a received RA would basically solve the issue for common
use cases.

The way I see it, there is a relatively simple interaction between a host 
and a CPE. Something we can understand. There a lot of weird stuff surrounding
access routers making it unlikely that all of that gets fixed. Even
in simple cases with static prefixes, access routers (and surrounding 
infrastructure) already do lots of weird things. Getting support for rolling
prefixes properly is an uphill battle.

At the same time, for IPv4, NAT will hide most of these issues. We can only
hope that CPE vendors will not add some kind of IPv6 NAT to deal with this
problem.