[ppml] IPv6 flawed?

On 9/13/07, mcr at xdsinc.net <mcr at xdsinc.net> wrote:
>   In IPv6, it is strictly not uncommon for a host to have multiple IP
> addresses.  So, you don't renumber the hosts or use NAT. You just let
> them have an IP address from your ISP, if you have one.
>   FURTHERMORE, shim6 will let you failover active connections from one
> host to another without starting a new TCP connection.
>
>   The problem is that, [snip]

The problem is that:

1. That's only close to true if you use stateless autoconfiguration
which suffers from such a severe security issue that it might well
drop out of use.

2. Its not actually true even if you do use stateless
autoconfiguration because you have to talk to other hosts on your
interior network via their full IP addresses, not via just the however
many bits that haven't changed. There's much more to communication
than a host's own IP address.

3. Shim6 is still in the research stage and at least 5 years from
ubiquitous deployment. Its not at all clear that it will ever make it
out of the research stage.


The answer might have been some sort of locality-mask that operates
like a reverse-netmask allowing interior hosts discard the high-order
address bits when communicating with each other. Since that wasn't
done and we're not on a path to IPv6 PI space, the answer is going to
be NAT.

Standard or no standard, someone is going to build it, someone is
going to sell it and when folks start buying it everybody else is
going to follow.

Regards,
Bill Herrin


-- 
William D. Herrin                  herrin at dirtside.com  bill at herrin.us
3005 Crane Dr.                        Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004