[arin-ppml] Discussion Petition of ARIN-prop-125 Efficient Utilization of IPv4 Requires Dual-Stack

Matthew Kaufman matthew at matthew.at
Wed Dec 29 12:55:48 EST 2010

On 12/29/2010 9:30 AM, Leo Bicknell wrote:
> It also illustrates the problem I have with the policy.  If we apply
> PP125 to the transfer market, I feel that we effectively nullify
> the potential usefulness of that market.  While I was never in favor
> of transfers because I did not think they would do enough good to
> outweigh the harm, I do believe that those who argued for it were
> standing on a solid footing.
> In this case I think PP125 is just too late to the party, so I'm
> in favor of sticking with the status quo.  That said, if we adopt
> PP125 I think it renders the transfer market nearly useless for
> it's intended goals,...
I agree completely with this analysis which captured exactly what I was 
about to point out. This proposal takes a very narrow view of "what 
additional public IP addresses are for", namely "adding public facing 
(web) servers" and "adding downstream customers". There's all sorts of 
networks which need public address space and for which *only* IPv4 is 
appropriate. Blocking them from getting IPv4 addresses from ARIN during 
the end-game is probably wrong, and blocking them from getting IPv4 
addresses via a transfer mechanism is definitely wrong.

As an extreme example, let us suppose the existence of the Distributed 
Legacy Computing Museum... a museum with dozens of facilities around the 
country in which computers and operating systems which never had and 
never will have IPv6 capabilities are on display and in operation and 
perhaps even reachable from the outside over IPv4 for interested 
researchers. What do they do when they acquire 100 additional surplus 
machines they wish to place on the network? (Answer: Get more IPv4 space 
from a transfer market or even a transfer in the form of a donation, 
perhaps. Or if they plan ahead, why not get them from ARIN next month?) 
If every IPv4 address they have now has a historical machine plugged in 
and operating associated with it, why wouldn't that be "efficient 

Similar cases exist for sensor networks that are in operation that want 
to add another hundred identical IPv4-only sensors over the next 10 
years and after the 10 years have a plan to switch to newer IPv6-capable 
sensors, etc.

Matthew Kaufman

More information about the ARIN-PPML mailing list