Re: [ICANN-EU] Re: [icann-candidates] IPv6

[lutz@iks-jena.de (Lutz Donnerhacke)]

* Andreas Bogk wrote:
>lutz@iks-jena.de (Lutz Donnerhacke) writes:
>>> Of course it will work. IPv6 contains intelligent network
>>> renumbering. With IPv6, renumbering will become common practice.

>> I hear your profund statement but I wonder if it will really work.
>
> Autoconfiguration works fine, even with multiple IP addresses.

Yep. I read the RfC yesterday again. They incooperated something like DHCP.
ARP is improved, too. It is definitly a great piece of work, but I still do
not understand, how to protect from false announcements (Felix will also
like to new snooping possibilities) and how to change addresses without
loosing connections.

Chapter 11 of RfC 2461 deals mainly with DoS and suggests to add the
orthogonal concept of the Authentification Header together with manual
configuration. Hmm... Orthogonal concepts are good. It is definitly an
improvement to the current practice, but we will observe practical use.
I fear uncontrolled autoconfiguration installations.

Chapter 12 does not deal with servers.

RfC 2374 seems to answer some of my open questions. First impressions:

While BGP4 is limited to 2**16 'real' providers worldwide, IPv6 is limited
to 2**13 (to 2**21) 'real' on called TLA. So my original sentence about IPv6
is still valid: How will those IDs be handled?

It could be registries only or (better for routing!) the major peering
partners. In order to ease routing, those IDs should be redistributed in
irregular intervalls to represent the current peering infrastructure. I do
not see a solution to manage this.

Automatic renumbering as mentioned in RfCs 2461,2462 does not cover the DNS
implications (Ever tried to change the IP number of root server? I did it)

But chapter 3.4 leave it open to use the TLA fields only for registries.
Hmm. Would ease a lot regarding renumberation, but cause a minimal (large)
routing table overhead.

Refering to RfC 1881 reallocation of address space is not allowed due to
routing issues.

So, I hope to climb back to the current level of IPc6 technology. You may
consider this message as a proof of my learning abilities. I might be wrong
on every topic, but I'd overcome my failtures.

> But I'm not sure Linux handles this properly yet.

I doubt it. Linux development is not as progressive as assumed. Most actions
are patchwork.