New ARIN ipv6 allocation policies

Iljitsch van Beijnum iljitsch at
Sun Sep 3 23:18:28 CEST 2006

On 3-sep-2006, at 20:09, Bjoern A. Zeeb wrote:

> there are other options with one of those be:
> filter out all prefixes of the PI block from the very first day and
> see what will happen who else will be and who'll complain and when.

Since presumably, everyone who has IPv6 connectivity, also has IPv4  
connectivity, this can be a workable approach. When you filter and  
you initiate a session, you should get back a "network unreachable"  
immediately and the client should fall back to IPv4 without delay.  
(This what was happened at the Montreal IETF because the IPv6 ISP  
there filtered the ARIN microallocations from which the IETF servers  
have their IPv6 connectivity.) The other way around isn't quite as  
smooth because then the remote end doesn't see any ICMP messages when  
it tries to initiate a session over IPv6 so it has to time out before  
falling back to IPv4.

> Why are people trying to put their "IPv4 world" directly into IPv6?

Conservation of mental energy? Also, many people working today have  
never worked in an environment with several layer 3 protocols like we  
had in the 1990s so they simply can't imagine anything other than IPv4.

> Why are people still doing multi-continent ipv6 tunnels to announce
> a prefix to give someone 'connectivity'?

Yes, those academic networks that think they can get by without paid  
transit in IPv6 and then dump everything they can't move over peering  
in those triple-continent tunnels are extremely annoying.

> We have some months left until we'll really urgently need to be
> able to switch everything over


We still have 1354.59 million IPv4 addresses to go and we seem to be  
burning some 165 million a year so yes, it looks like current IPv4  
practices can continue for a few more years.

More information about the ipv6-ops mailing list