September 13, 2004

Global Addressing Policy

I see that ICANN's ASO - a body composed mainly of the regional IP address registries (RIRs) - has submitted a document entitled INTERNET ASSIGNED NUMBERS AUTHORITY (IANA) POLICY FOR ALLOCATION OF IPv4 BLOCKS TO REGIONAL INTERNET REGISTRIES

It is an interesting document.  And I won't do more than mention in passing that it was written mainly by those who receive the allocations described by the policy.

As for the policy expressed in the document:  There is no doubt in my mind that the policy itself as articulated in the document is a rational one and appears based on the lessons of years of RIR experience.

In terms of its impact on the overall internet this policy is of much greater import than all of ICANN's DNS policies and DNS task force reports and UDRP's put together.

Nearly all of ICANN's DNS impositions can be bypassed simply by innovating at a lower level of the domain name hierarchy.

However, IP addresses are the sine qua non of existence on the internet; a computer is not on the internet unless it has an globally valid and routable-to IP address (or is behind a NAT that has one.)  If a region, a country, an ISP, a business, or a person can not get a usable number of addresses, then they are not really on the internet.  In some future political or economic competitive situations the winner may well be the entity that has better access to IP address resources.

There are two points about the document that strike me as interesting.

First is that is purports to be an IANA policy.  Where is the adoption of this policy, by IANA or ICANN's board playing stand-in for IANA.  Or is this merely a suggested policy?

Second is that this document is giveth-only, it has no taketh-away.  This policy does not seem to envision any recovery of addresses by IANA from the RIRs or any re-allocation of address blocks.

My last conversation with Jon Postel was on exactly this issue of recovery and re-allocation.  The RIRs make sense as an allocation mechanism because they coarsely correspond to chunks of internet connectivity.  In other words, the connectivity within a region covered by a RIR tends to be more intensive than the connectivity between regions covered by different RIRs.  (As I said, this is merely a course correspondence and is being increasingly leavened by concessions to regional pride.)  As the connectivity of the internet evolves over time the modularity of today's connectivity may change and the efficient aggregation of address blocks might then suggest an equally evolved RIR structure.  Jon and I were in agreement that the RIR system and the systems of allocations made to RIRs ought not to be considered immutable but rather to be considered as flexible and subject to periodic re-evaluation.  This policy seems as if it leaning towards immutability rather than towards flexiblity.

Posted by karl at September 13, 2004 12:08 PM