In message <[email protected]> Michael Thomas writes: > On 07/30/2012 04:08 PM, Michael Richardson wrote: > >>>>>> "Michael" == Michael Thomas <[email protected]> writes: > > Michael> If we believe that ipv6 is ready to go for mass > > Michael> deployment, why do we > > Michael> not pressure home router vendors to default to sending > > Michael> router advertisements > > Michael> with ULA addresses that, if necessary, get NAT'd at > > Michael> the border just like > > Michael> 192.168 space does today. > > > > Michael> I mean, nothing bad would happen, right? > > > > Do you mean, NAT64'ed? > > > > > > Yes, just nat v6 internal addresses to v4 on the isp side when you > don't have v6 connectivity. If you're reading this and inclined to > a NAT rant, you probably didn't understand my post. > > Mike
Mike, There would be no advantage to either the user or the service provider to allocate from IPv6 PI space rather than IPv4 PI. Whether the user has a 10/8 with 16 million hosts, or a /64 with 16 billion billion (whatever that is called), makes no difference. The provider needs to allocate exactly one IPv4 address if we are doing NAT64. If the whole neighborhood is connected as in a consumer ISP, then they allocate a large pool. Last I was on one of those, the pool was a /20 with 4,906 IPv4 addresses. It doesn't take long to run out of IPv4 addresses that way. And we pretty much have done so. Allocating IPV4 PI addresses on the client side and IPv6 on the service provider side doesn't help either. The client could only address IPv4 content providers and to make a IPv4-in-IPV6 address at least one routable IPv4 address is needed on the NAT. If the NAT still needs one IPv4 address, nothing is gained on the provider side (where the IPv4 address depletion resides). If you see some advantage that solves the IPv4 address depletion (a big point of the transition to IPv6 exercise), then I've missed it. If so, please point out what I missed. If both side, client and provider, are IPv6, then there is no reason to bother with a NAT at all. The router doing the NAT has to know its own IPv6 routable address. That address has to be configured or learned somehow such as via DHCP. There is nothing to be gained by handing out a /128 to the router for it to NAT, rather than handing out a routable /64 and letting the customer router allocate routable addresses. There are 4 billion times as many IPv6 /64 available as there are IPv4 addresses so NAT is unnecessary. I hope this wasn't perceived as a rant about NAT. Curtis _______________________________________________ homenet mailing list [email protected] https://www.ietf.org/mailman/listinfo/homenet
