[address-policy-wg] RE: Question

Hi,

On Thu, Apr 20, 2006 at 01:08:20PM +0400, Maxim V. Tulyev wrote:
> > (thanks for doing proper quoting - which is *below* *properly trimmed*
> > original articles)
> This sticks to me as the good FIDONet behaviour ;)

Thanks.

> > > it does _NOT_ work for IPv6 in the wild, by the way.
> > Well, YMMV, but my customers claim "it does work" - with IPv6, and in
> > the wild.
> 
> I tried to announce network 2001:4058::/48, and in fact, I couldn't get 
> working connectivity.
> 
> Of course, if you have two channels, and one of them also announces 
> entire /32, it will be "seems to work", because you will get incoming traffic 
> from there. 

Yes, that's the underlying assumption when using "PA slice multihoming" -
the aggregate is always visible in the global table.

> But if that channel fails, you will lose connectivity at all. 

... to destinations that filter the /48 and have no default route.

Which is not so much different from "if one of your upstreams is messing
up their routing completely, you'll have problems reaching parts of the
internet".

There is no way anyone can guarantee reachability to any place at all
times.

> So, why? Local peerings? Something else?
> 
> That's NOT like PI or more specific IPv4 that can be announced and is working 
> in the wild as an independent part of Internet.

As long as the aggregate is visible, reachability is as good as for IPv4 PI
space, if not better.

Gert Doering
        -- NetMaster
-- 
Total number of prefixes smaller than registry allocations:  88685

SpaceNet AG                    Mail: netmaster at Space.Net
Joseph-Dollinger-Bogen 14      Tel : +49-89-32356-0
D- 80807 Muenchen              Fax : +49-89-32356-234