I was reading through some of the notes from the talks at NANOG, the North America Network Operators Group. In particular, notes about the transition to IPv6 caught my eye.

I’ll be the first to admit I don’t know enough about the protocol, but what always struck me as odd was that there’s no multihoming for enterprise customers. Only carriers get to multihome (advertise their prefixes out multiple providers), and by strict interpretation anything longer than a /46 is to be dropped (the default allocation for companies is /64).

The talk went over some of the barriers to IPv6 adoption, and one of the reoccuring themes was lack of multihoming. For some companies, this is essential since the Internet is their only revenue stream. So maybe it’s not that multihoming isn’t available, it’s just that there is no good high availability solution for enterprises, and in our IPv4 mindset we think of multihoming.

The argument against this is that it goes against aggreagation. If a network is to be multihomed, it can’t be aggregated, and there’s one more prefix out on the Internet. Already on the Internet we have almost 180,000 prefixes on the Internet. Increase the address space by a factor of 2^96 and routers will be overwhelmed.

The counter argument to that is largely that it’s possible to build routers that can outpace the growth of the Internet routing tables.

Regardless of the arguments, I’m happy to see that groups are taking a pragmatic view of IPv6 adoption, and looking to overcome the hurdles.