[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Tunnel-to-NAT scenario



On 17 jun 2008, at 13:47, Jari Arkko wrote:

I hope no one is no longer arguing that we should do a double NAT. I think that would be wrong because tunneling ensures that your IPv4 traffic is as intact as possible. Yes, it may eventually get NATted, but I do not want to create technology that forces you to do it no matter what is at the other end. For instance, with a tunnel-based crossing of the v6-only cloud customers sites that have public IPv4 space will not see any NAT effects.

I agree that in the case of passing IPv4 through IPv6 without NAT tunneling is better. However, the case that we're discussing here is the one where there is also NAT involved. In that case, pretty much all the fields in the IPv4 header are mangled/ignored anyway, so there is no real value in carrying them across the IPv6 cloud intact.

Basically, by borrowing 32 bits of the IPv6 destination in the IPv6 header, we can compress away the entire IPv4 header because none of the fields in there are needed later.

By going one step further and also adjusting the transport checksums it's possible to use the same NAT64 that also serves IPv6 hosts that want to talk to IPv4 destinations, which saves implementing two sets of translators or translators that support two different types of translation.