[arin-ppml] Just a reminder of some quick mathematics for IPv4 that shows the long term impossibility of it
tedm at ipinc.net
Fri May 13 11:40:25 EDT 2011
On 5/13/2011 6:30 AM, William Herrin wrote:
> On Thu, May 12, 2011 at 7:44 PM, Owen DeLong<owen at delong.com> wrote:
>> On May 12, 2011, at 3:46 PM, William Herrin wrote:
>>> On Thu, May 12, 2011 at 4:23 PM, Ted Mittelstaedt<tedm at ipinc.net> wrote:
>>>> So how exactly do we get the other 4.5 billion people on the Internet
>>>> using IPv4?
>>> Survey says: NAT.
>> That does not put the other 4.5 billion people on the internet.
> The half billion or so who've joined the Internet behind NATs this
> past decade seem to think differently. Who am I to disagree with them?
> Who are you.
I think your just seeing if we would take you serious, Bill. You
cannot possibly believe the Internet can add all those additional
people in the same manner as they have added the first 2 billion.
NATs work by using an external public address. When all the
external public addresses are gone, you can't setup any more NATs
Perhaps if 10 years ago carrier-grade NAT had been mandated and
all public addressing had been prohibited from end-users, and only
allowed to operate on transit router and networks, while all end-node
AS's were just given a small allocation, like a /24, and expected to
put all their end users behind centralized NATs, why then it might
have slowed address uptake to where it would have been feasible.
But that didn't happen because the CPUs and hardware of the time were
not powerful enough to do it for the core to do it. And it would
have required giving up a fundamental property of the Internet which
is that every host have a unique IP address.
All that happened with NAT is that we changed the premise slightly to
allow for the notion that small pockets of hosts on the Internet
could have unique IP addresses from other small pockets of hosts,
mixed in with a sea of regular hosts that had unique addresses, but
with the stipulation that those small pockets could never be coalesced
into significantly large pockets. That slowed IP address uptake and
gave us the breathing room for large behemoths like Microsoft and
Cisco to fully implement IPv6 in their products, but the other part
of that fundamental tradeoff - the stipulation that the small pockets
cannot be coalesced - now prevents NAT from getting any larger.
For anyone who really believes NAT can get the rest of the 4.5 billion
people in the world on the Internet without significantly changing
anything for the first 2.5 billion, consider this. The entire Internet
would have to be so disrupted to put a carrier NAT solution into place,
with so many people having to give up routable numbers, that it would be
equivalent to the disruption caused by dropping IPv4 and going to IPv6.
The resulting solution, with translators everywhere and significantly
increased expenditures for all networks in hardware,
increased fragility caused by introducing so much additional
hardware into the network, and decreased flexibility as simple packet
moving on the Internet turns into a huge bureaucracy, would be a
far worse result than a single routable network with full end 2 end
connectivity available to any host that wants it, a network where
routers are routers and hosts are hosts and network devices go back
to being simple devices in the dusty corner, plugged into printers and
For everything there is a season, and a time for every purpose under
heaven. IPv4's time is ending, IPv6's time is beginning.
More information about the ARIN-PPML