BGP over 2002::/16

Iljitsch van Beijnum iljitsch at muada.com
Tue Dec 18 13:35:28 CET 2007


On 17 dec 2007, at 21:13, Jeroen Massar wrote:

>> I finally managed to get my test hardware set up. This is a Cisco  
>> config
>> that lets you peer across 6to4:

> Cool! I am totally surprised that it works though, but theoretically
> indeed why wouldn't it. Maybe the KAME & Linux stacks should be  
> updated
> to support this too. Thanks for testing this Iljitsch!

Well, the real question is: would a setup like this be useful as a  
mechanism to get peering between not-so-well-connected IPv6 networks  
off the ground? I think it could work very well but it seems a bit  
klugy.  :-)

An alternative would be to set aside a non-6to4 address range but use  
6to4 tunneling over that. I don't think Cisco, BSD or Linux hardcode  
2002::/16 to 6to4 interfaces. This way, the BGP routers are less  
exposed to random 6to4 traffic from elsewhere on the internet.

Anyway, here's the config on my FreeBSD 5.x / Zebra box that is the  
other side of this link:

# ifconfig stf
stf: flags=1<UP> mtu 1280
         inet6 2002:5395:4101::1 prefixlen 16

!
router bgp 12854
  bgp router-id 83.149.65.1
  neighbor 2002:52c0:5a1a::1 remote-as 9000
  neighbor 2002:52c0:5a1a::1 ebgp-multihop 255
  neighbor 2002:52c0:5a1a::1 update-source 2002:5395:4101::1
  no neighbor 2002:52c0:5a1a::1 activate
!
  address-family ipv6
  network ::/0
  network 3ffe:2500:310::/48
  neighbor 2002:52c0:5a1a::1 activate
  exit-address-family
!

And more evidence that it works:

12  if-6-0.core1.ad1-amsterdam.ipv6.teleglobe.net  46.942 ms  59.032  
ms  45.915 ms
13  ix-2-0-496.core1.ad1-amsterdam.ipv6.teleglobe.net  60.27 ms   
46.543 ms  46.397 ms
14  2001:1af8:1:15:205:5fff:fe18:307f  47.526 ms  46.898 ms  46.844 ms
15  2001:1af8:2:5::2  46.912 ms  46.954 ms  46.806 ms
16  2002:52c0:5a1a::1  62.225 ms  61.432 ms  65.976 ms
17  2001:1af8:5:1:204:27ff:fefe:249f  65.351 ms  65.932 ms  65.245 ms



More information about the ipv6-ops mailing list