> Date: 24 Nov 1996 11:40:00 +0200
> From: Kai Henningsen <kai@khms.westfalen.de>
> To: submit-linux-dev-kernel@ratatosk.yggdrasil.com
> Subject: Re: IPv6 and the "average user"
>
> ecki@inka.de (Bernd Eckenfels) wrote on 23.11.96 in <577j5k$2bc@nz12.rz.uni-karlsruhe.de>:
>
> > Kai Henningsen <kai@khms.westfalen.de> wrote:
> > > To keep them alive as long as possible, routes need to have long prefixes
> > > - or, to put that differently, providers have to share one or very few
> > > prefixes among all their customers.
> >
> > > And *that* makes for too few IP addresses for every customer.
> >
> > Does that mean with IPv6 Address Blocks will be assigned to ISP and the
> > Customer has to switch Networks if they change Providers?
>
> Yes. No way to avoid that has been found; instead, the IPv6 people have
> concentrated on making it as painless as possible.
Suppose the "customer" is an ISP themselves? Suppose they want to be
multihomed? What then?
For that matter, who decides what organizations qualify as "ISP"s and
get assigned blocks?
> > > However, you'll still need to change your addresses when you switch ISPs,
> > > unless you are big enough so the backbone routers will be willing to make
> > > routing table entries just for you. Because of this, renumbering a network
> > > is a lot easier in IPv6 - each box *must* support being told their address
> > > by the router. At the router, just configure the new 80 bit prefix. The
> > > rest is automatic. (Well, if you do DNS, you'll have to change that, too.)
> >
> > Ah yes.. thats nice... Thanks for that info.. is there a RFC especially for
> > that policy?
>
> For which policy?
>
> MfG Kai
>