Subject: Re: v6 (was Re: -current sendmail cancer in IPv4-only kernel)
To: None <current-users@netbsd.org>
From: Feico Dillema <dillema@acm.org>
List: current-users
Date: 05/09/2000 11:43:50
On Mon, May 08, 2000 at 06:01:50PM -0700, Jonathan Stone wrote:
> 
> In message <523dns1x2t.fsf@sean.ebone.net>, Sean Doran writes:
> >What?!?!?!?!  It's Just CIDR, precisely as in IPv4.  And
> >with all the pluses AND minuses of CIDR.  There is
> >absolutely nothing different about routing IPv4 and IPv6.
> >Unfortunately.
The routing still works the same way, yes. But the IPv6 addressing
plan is designed to be hierarchical and aggregatable. There likely will be 
violations to this hierarchy for a limited amount of big networks in
order to avoid/ease renumbering events etc, but there's a strong sense
in the IPv6 community that these routing hierarchy violations should
be as rare as possible to keep routing table small.

> As best I undrstand things, you can't do that with IPv6: Instead one
> has to multi-home all hosts with multiple addreses, one from each ISP.
> Failover after a back-hoe takes out one ISP becomes, uh, interesting.
> This is a very real concern to some people.
Multi-homing like that is still one of the issues that need (and gets)
attention. In a way with IPv6 even more complexity is
shifted from the core of the network (routers) to the end-systems or
at least the edge of the network. 

> But other than that: yes, you're right. IPv6 has the pros and cons
> of CIDR.  Renumbering, for example.
Yes, but that is also the reason why autoconfig and autorenumbering
protocols have gotten so much attention for IPv6. In IPv6 we may be
able to live with the occasional renumbering, were such a thing would
give nightmares to network admins in IPv4.

Feico.