Re: draft-bourbaki-6man-classless-ipv6-00

[Tom Herbert <[e-mail]>]

On Sun, Jun 4, 2017 at 6:22 AM, Peter Hessler <[e-mail]> wrote:
> On 2017 Jun 04 (Sun) at 22:11:59 +0900 (+0900), Lorenzo Colitti wrote:
> :On Sun, Jun 4, 2017 at 9:48 PM, Peter Hessler <[e-mail]> wrote:
> :
> :> :There is no workable solution if you want to assign 64 bits to the bus, 64
> :> :bits to the routing system, and M bits for mobility, unless M = 0.
> :>
> :> That's *exactly* why bus and routing MUST NOT require a specific size.
> :>
> :
> :If you don't require a specific size "operational consistency with IPv4"
> :will inevitably lead to one /128 per device and NAT. As a host developer I
> :don't want to implement NAT, because it's bad for my users.
>
> One address per device *is a good thing*.  From the OS perspective,
> having to select which ******** address to use as the source IP is
> horrible.
>
It wouldn't be particularly difficult to implement random address
selection in an OS, but I imagine that might confuse some applications
that assume there is only one address in the system. However, for the
purposes of address obfuscation I don't see why a host would need 2^64
addresses, a few billion addresses should suffice.

> NAT means two things:
>
> 1) You can't depend on directly accessing the device.  This is true with
> Firewalls, so you have to consider that anyways.
>
> 2) You can't embed the IP address inside the packet data.  That's a
> stupid idea, and shouldn't be done.
>
> Besides, *Lots of people are already deplyoing IPv6 NAT in the wild*.
>
NAT is bad because it requires networks nodes to maintain and track
connection state (like stateful firewalls which also need to go
away!). It only supports certain protocols and extension headers, it
has motivated the use of NAT keepalives in UDP which are nothing but
junk packets, and for connections all packets must go through the same
network device which becomes a bottleneck and prevents multi-homing.

Tom
>
> --
> You will feel hungry again in another hour.
>
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> [e-mail]
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------