[ppml] Proposed Policy: IPv4 Countdown

[william(at)elan.net]

On Fri, 16 Mar 2007, Stephen Sprunk wrote:

> We do not have a choice.  The IPv4 address space _will_ be exhausted, and
> it'll happen in about four years if we do nothing.

4-5 years are projections when IANA may not be able to fulfil request by
RIR for new /8 which I personally think would be closer to 6. Most RIRs 
request space in advance and have some space in reserve up to 2 years
and there is also space unfilled from legacy class-b blocks that RIR at 
that point will most likely start to use all together.

Real end of available space will depend on RIR and likely to be 2014-2017.
Of course estimates vary based on who you ask and some people do have
interest in making numbers appear lower then they are too.

> The best projections, by people who are quite authoritative on the 
> matter, is that reclamation will buy us six more months.

It will "buy" more time - closer to 3-4 years probably. If you're looking 
for numbers, currently 65% of Class-A,B,C (1.0.0.0-223.255.255.255) is 
allocated and of that 45% is routed leaving 20% as not routed (which has 
some amount in use internally but I'd be surprised if its even 1/4th
of that) which means potentially equivalent of 46 /8 blocks could be
reclaimed right now (and those numbers will grow too). But if we want
do do reclamation such decision better be made soon and process start
as well.

For reference where I calculate the numbers, see
  http://www.completewhois.com/statistics/ip_statistics.htm

> claim, as you did in another message, that it'll really buy us 5-10

definitely not 10 years but 5 is in theory possible if more currently
used blocks go dark as well as if reclamation effort includes not only
"dark" blocks but also blocks that are lit but that are not seriously
in use (but that I mean asking large companies, universities and
military that early on got /8s and smaller companies that still hold
/16 but are now like most others use NAT to renumber and return most
of the /8 or /16).

> years, you need to come up with better studies and data than we already 
> have.  And we'll be in the same boat again after that much time anyways, 
> so we might as well save the effort and convert now before we buy/deploy 
> _another_ hundred million routers and PCs.

The real question is would we have situation 7 years from now that
most enterprises and PCs are capable of handling IPv6 and that we can
start just assigning IPv6 to all in addition to IPv4. I think that
it will happen even though that last few years would be rather for
painful those in IT (although for consultants it can be rather good 
time...) when everyone who cares and wants to be on the internet
has to enable it.

Another guess is that there would be software option on the routers
& firewalls to do NAT-like translation to IPv4 for those only on IPv6
and the other way around and that it would be done on the routers and 
firewall of devices that currently do NAT translation at many enterprises 
(NAT is actually quite popular at the edge despite efforts by many to
say that its bad...) - I'm guessing that multitude of those NAT router 
vendors that consumers and small businesses buy would actually use end
of IPv4 as way to encorage consumers to buy such new device.

-- 
William Leibzon
Elan Networks
william at elan.net