[6bone] RFC2772 rewrite

[Roger Jorgensen]

At 06:59 PM 11/12/2002 +0100, Pim van Pelt wrote:
><snip>
>* Should we seperate the 6BONE cloud from the production IPv6 cloud ? This
>is not really an RFC2772 issue, but does come to mind.

It isn't really a RFC2772 issue but I guess but it's time to start thinking 
and
have some more focus on all the routing issues we see in IPv6 space so
we can make it a RFC2772 issue?:)


About what Pim said, it's not a question about IF. I would say it's more a
question on HOW.
The reason for that statement are that there are people out there that simply
don't trust the routing in the 6bone cloud. Are too many bad tunnels and 
insane
tunnel connections out there. I've killed many tunnel/transit connection just
because the connection are very bad.
With time will we kill all tunnels to our ASN and run 100% native. With that
move it would be logical to also consider,
         "how do we want to keep the connection to 6bone?"
         "do we want it?"

About the routing mess issue, all connected parties to 6bone need to stop
up and consider,
         "do we have good enough infrastructure to provide transit 
connectivity?"
If you can't provide transit, sit down with your routing and start filtering.
It's time to stop providing free transit to everyone and accept it from 
everyone,
6bone and the IPv6 Internet (yes I see them as two separate network actually)
have grown too big and too complex for the free exchange of transit to work
anymore... the ghost routes are one good example.


Pekka's document, 
http://www.ietf.org/internet-drafts/draft-savola-v6ops-6bone-mess-01.txt
are a good guide on the routing issue.



Some experience I've had with IPv6 routing, guess I'm not the only one with 
it??
I wanted to reach a site I knew was in Europe. I got a route to it from EU, 
to US,
then going around in US before it went back to EU and then finaly hit the 
target.
Pingtime was stable 3-400ms...with decent filtering I probably could have 
reached
the same site within 200ms.
What is worse are that there are some ASN's out there that are "hijacking"
routes, or that's how it feels like when you peer with that ASN.
I can have two peering, one with AS1 (hijacking ASN and 100ms away) and one
with AS2 (a 10ms tunnel), for some strange reason have I seen it again and 
again
that the route are ALWAYS going over AS1. I've heard the same thing from 
others
to so it's not only me that are seeing this.





---
Roger Jorgensen (rjorgensen@upctechnology.com)
System Engineer @ UPC Technology / IP engineering
handles: ROJO1-6BONE ROJO9-RIPE RJC10-NORID