Re: [v6ops] ULA draft revision #2 Regarding isolated networks

[Tim Chown <[e-mail]>]

On 28 May 2014, at 09:09, Philip Homburg <[e-mail]> wrote:

> In your letter dated Wed, 28 May 2014 16:24:59 +1200 you wrote:
>> I understand the problem. Users are idiots, including us when we
>> act as users. I am not arguing that ULAs will prevent all operational
>> problems caused by private addressing. I am only arguing that there will
>> be private addressing and that ULAs will cause fewer problems than
>> RFC1918-for-IPv6 would.
> 
> Maybe we can keep a bit of context around what scenario we are talking about.
> 
> Say:
> 1) Small organisation, doesn't get PI space, has to renumber when moving to a different
>   ISP. The IPv4 solution is to use NAT. Maybe we can do better for IPv6. Probably
>   collisions of ULA space during a merger are not a big deal.
> 2) Bigger organisation, does have PI space and has a more or less homogenous network.
>   No real need for ULA.
> 3) Big organisation, wants to have a secret network that only communicates through
>   application layer gateways. Should be able to pick a proper ULA if they need one.
>   Otherwise, though luck. No real need for ULA. It may make source address selection
>   a bit easier, but that is only a small part of the puzzle.
> 
> I think the only interesting case is 1). Whether or not we make renumbering easy enough
> that they don't do NAT is an open question. A large part of that is in software
> configuration: is it possible to store the current prefix in such a way that software
> can easily pick it up.

And we spent two years batting that topic around in 6renum, for site/enterprise networks (not targeting homenets).

IPv6 network renumbering on the wire = reasonably straight forward (Fred’s work is now 10 years old)
The complexity is all the places addresses are embedded in interesting ways, in your sites and sites you speak to.

Three RFCs emerged, see http://datatracker.ietf.org/wg/6renum/

Tim