[6bone] Re: routing concern

[Christian Schild]

Am Mittwoch, 31. Juli 2002 23:39 schrieb Petr Baudis:
> Dear diary, on Wed, Jul 31, 2002 at 11:03:55PM CEST, I got a letter,
> where Ronald van der Pol <Ronald.vanderPol@rvdp.org> told me, that...
>
> > On Wed, Jul 31, 2002 at 22:01:07 +0200, Petr Baudis wrote:
> > > And as the time goes on, people obviously tend to sacrifice the
> > > tunnel peerings for native ones and the reliability improves. The
> > > natural process.
> >
> > This is what we have been trying to do for the last couple of
> > years, but without much success. The 6bone is still too unstable.
> > Look at the OS mailing lists. End users are disabling IPv6 in
> > their OS because it does not work.
> >
> > We don't need a stable IPv6 network tomorrow. We need it today.
> > I doubt if we can make the 6bone stable very soon.
>
> Again, this is not problem of 6bone, but problem of its users. They must
> move from tunnel peerings to native peerings, and then they will also move
> from 6bone. When you will cut out 6bone, you will just force them to move
> all their tunnelled links etc to 2001::/16 space, polluting it terribly,
> making it impossible to distinguish Bad Peers and Good Peers easily. And
> you will just make situation worse in another part of IPv6 internet.

As others pointed out, stability has nothing to do with native or tunneled
connections. The point I want to make is, that we have no time to wait 
for the 6bone to become stable, be it either by evolving native peerings or 
by educating it's users how to filter and to route. 

In fact, I don't believe it is possible at all to stabilize the 6bone. It is 
marked 'experimental'. People come and go in the 6bone, new players will
do nasty stuff as 'it is only experimental'. Black holes will never go away, 
because new people want to 'learn' in the 6bone, how to run an IPv6 network. 
Learning implies making mistakes. 

Thinking of this, I believe it is vital to detach the production network 
from the 6bone _now_, and it is also vital to show everyone whats the 
difference.

'You have an 3ffe:: prefix? well, then don't expect anything.' 
'Yes, 2001:: should be better, but we can't guarantee that our (production) 
peering partners, won't use routes through the 6bone.' 
The second sentence _should_ say 'Yes, 2001:: is better, it's production'.
To achieve this I would like to urge every ISP not to mix up their 2001:: 
prefix with 3ffe::, and to encourage them to detach it from the 6bone. 
If this would become a general rule how to play in the 6bone, I believe a 
stable production network is possible. 


Their is one drawback. 
Todays IPv6 network _is_ the 6bone. If I want to reach someone, I often 
have to traverse the 6bone. So most times I like to consider the 6bone 
as the 'IPv6 backbone'. I want to use a 'productive' route, but sometimes 
only a route through the 6bone is possible. It's a kind of last resort 
for reachability. 

Complete detachment of the 6bone is not possible today, as we lack of 
productive connectivity. But we could use the 6bone as seldom as possible.

Considering an ISP with a 6bone node and one or more peerings to a 2001__ 
neighbor, that would require some preferences set for a productive route 
from the productive neighbor, or a lower preference for routes coming from 
the 6bone. Then the ISP shouldn't accept or announce 3ffe:: routes to its 
peering partners and should only use his own 6bone node for 6bone 
connectivity. This way the ISP could reach every possible IPv6 network and 
if possible I will use a route through my productive peerings. Ok, my peering
partners still could route through the 6bone, but if he uses the same rules
like me, that won't happen. 

This method may only be a starting point for discussion. I'm sure there 
are other ways to detach the 6bone (partly) from the productive network. 


Christian

-- 
JOIN - IP Version 6 in the WiN  Christian Schild
                 A DFN project  Westfaelische Wilhelms-Universitaet Muenster
Project Team email:             Zentrum fuer Informationsverarbeitung
          join@uni-muenster.de  Roentgenstrasse 9-13
http://www.join.uni-muenster.de D-48149 Muenster / Germany
email: schild@uni-muenster.de,phone: +49 251 83 31638, fax: +49 251 83 31653