[arin-ppml] End non-public IPv4 assignments?

[Jack Bates]

On 1/25/2011 11:20 PM, Owen DeLong wrote:
> On Jan 25, 2011, at 8:56 PM, Jimmy Hess wrote:
>
>> All the networks that do not connect publicly must do  is negotiate
>> between themselves in regards to which addressing will be used by
>> which network.
>>
> This includes the following (incorrect) assertions:
>
> 1.	These networks know about each other's existence.
> 2.	These networks have a viable means by which to carry out said
> 	coordination.
> 3.	These networks are sufficiently limited in scope as to make such
> 	an activity feasible.
>

To expand upon Owen's explanation, these networks aren't built 
automatically connected. They form and drop connections with multiple 
entities over time. They cannot use NAT translation between the networks 
due to the nature of the communications they do allow across the 
peerings (there are always firewalls between the interconnecting 
networks). To make matters worse, these private networks will most 
likely be the slowest to be able to adapt to IPv6.

Just imagine how many networks might privately peer with walmart, 
shipping/trucking networks, networks that handle tracking of trucks on 
the road, credit card processing networks, etc. How they peer is 
irrelevant. Some may peer via slow T1 circuits while others my peer with 
VPN. There is a huge complex set of dynamic internetworks which require 
unique addressing. IANA via the RIR's is the only mechanism for insuring 
these unique addresses to allow the dynamic nature of such an internetwork.

Let's talk SIP. A lot of telephony has moved into SIP, from ONTs and 
DLCs talking SIP to local exchanges to international voice traffic. How 
the traffic gets from point A to B is irrelevant. What is important is 
that the communication requires unique addressing even when it is vpn or 
private peering.


Jack