Re: graceful renumbering of CPE networks

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

On 2019-02-21 04:25, Michael Richardson wrote:
> 
> Lorenzo Colitti <lorenzo=[e-mail]> wrote:
>     > FWIW, I think the reason that the only way to renumber IPv6 networks is
>     > (mostly) there is no way to explicitly communicate the user prefixes from the
>     > CPE to the network. I think most deployments inject routes based on DHCPv6 PD
>     > snooping, and the implementations of that generally only support one injected
>     > route per DHCPv6 client.
> 
> on the specific comment:
>   BMG (PPPoE) devices usually terminate the DHCPv6 at the BMG, and so the BMG
>   sees the radius packets that went into forming the PD, so it's not DHCPv6 PD
>   snooping, exactly.  There is nothing in PD or radius that prevents multiple
>   prefixes from being delegated, it's just a software problem :-)
>   For the cable-modem situation, the scenario is very different.
> 
> I was confused at first by your comment, so I read it a few times.
> I think you are suggesting that if we could have multiple prefixes (the old
> and the new), that the renumbering event would no longer be a flash
> renumbering.  That the mode of operation that 7084 recommends essentially
> results in flash renumbering.  There is no apparent way to introduce a new
> prefix while the old prefix is still alive.

Whereas of course RFC 4192 specifically describes planned renumbering
(for enterprise network networks with actual humans in charge.)
However, it seems to me that RFC 7084 is aimed mainly at manager-free
scenarios. I do agree that a document on manager-free renumbering
might be appropriate, but I think it's more than a simple fix to 7084.

Mumble HNCP mumble. RFC 7788 doesn't really seem to tackle the issues
we've been discussing here.

    Brian

> 
> The lack of old/new prefix definitely seems like a bug and it looks like
> something that 6man ought to fix with an update to 7084, ideally in a
> standards track document.
> 
>     > On Wed, Feb 20, 2019 at 4:46 PM Ole Troan <[e-mail]> wrote:
> 
>     > Top posting. Two points here.
> 
>     > 1) Flash renumbering breaks stuff. Especially for people running servers.
>     > Allowing people to host and own their own content in their own networks
>     > would do a lot more to protect people’s privacy rather than changing
>     > addresses. So let’s agree that the problem you want to solve is a
>     > misconfiguration/error case, not recommended operational practice.
> 
>     > 2) We should always strive to make mechanisms more robust. Your suggested
>     > point solution already have alternative broader solutions. E.g. MPMH host
>     > or added rules in SAS. It might be worth looking at the more general
>     > problem.
> 
>     > Ole
> 
> --
> Michael Richardson <[e-mail]>, Sandelman Software Works
>  -= IPv6 IoT consulting =-
> 
> 
> 
> 
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> [e-mail]
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------
>