24th APNIC Open Policy Meeting - APNIC plenary: The future of IPv4

Thursday 6 September 2007, Intercontinental The Grand Hotel, New Delhi, India


Meeting commenced: 09:15 am

Akinori Maemura
Paul Wilson
Randy Bush
Geoff Huston
Kusumba Sridhar
Tulika Pandey

Paul Wilson opened the session and discussed some housekeeping issues.

The Chair welcomed the attendees and introduced Talika Pandey from DIT, who presented the opening statement.

  1. Welcome address
    Future of IPv4 - India perspective

    Tulika Pandey, DIT


    The presenter discussed the Indian government perspective on the future of IPv4.

    It was noted that the Indian government is concerned about the slow uptake of IPv6.

    The government applauded the efforts of ISPAI and SANOG in showcasing the Indian IT industry at this conference. It was also noted that a very important event, IGF, would be held in India in 2008. It was noted that India is a leader in IT, and that expectations were high.

  2. IPv4 unallocated address space exhaustion

    Geoff Huston, APNIC


    The presenter noted that this issue has been studied for some time, because almost since the inception of IPv4 there has been an understanding that eventually the address pool would be exhausted.

    The presenter discussed the current status of IPv4. It was noted that there are 46 /8 blocks left in the unallocated address pool. It was also noted that around two thirds of allocated blocks are being advertised in BGP (the public Internet), and an additional 21.5 /8s were inside the RIR pool (not all of which are usable). That leaves 25 /8s unaccounted for.

    The presenter discussed the fact that change is a long-term, incremental process. He also noted that past behaviour could often be an indicator of future behaviour. It was noted that the good consumption data starts at 1990, but the data prior to that is not as clear.

    The presenter discussed the advent of classless addressing, which occurred when there was concern in the early 1990s about pool exhaustion. It was noted that the RIR system imposed a much more conservative view of addressing and was much more attuned to actual need. It was noted that the Internet has grown rapidly during the past three years. It was noted ARIN, RIPE and APNIC have allocated a similar number of addresses, while LACNIC and AfriNIC have allocated only a fraction of those addresses.

    The presenter noted that the current system is conducive to encouraging advertising of addresses.

    The presenter stated that his modelling technique (which is based on the assumption that past behaviour dictates future behaviour) projects that the unallocated pool will be exhausted by approximately 22 April 2010. It was noted that this projection also assumes that there will be no panic, no change in policies or demand dynamics, no disruptive externalities, rationing, withholding or hoarding. Therefore, it was suggested that the politics of scarcity would create a critical situation within the next one to two years.

    The presenter discussed possible scenarios for the post-exhaustion period, including NATs, address markets, routing fragmentation, and IPv6 transition. The presenter discussed the advantages and disadvantages for each scenario.

    It was noted that long-term use of NATs is impractical. It was also noted that the costs of IPv6 transition do not match the business case for service providers, and it was now too late. Thus, a long and protracted dual stack period is now inevitable.

    The presenter noted that RIRs are now discussing how to ration IPv4. Ultimately, however, this does not matter, as the big issue is what will happen after exhaustion.

    The presenter noted that the broadcast spectrum distribution process was once an administrative process, but as the resource became scarce market mechanisms took over. It was suggested that these forces would also apply to IPv4 addresses.

    It was noted that the elements of the current system that work, such as routing, applications, fuelling Internet growth, and integrity of network architecture, need to be preserved.

    The presenter asserted that we must consider implications of choices in a global sense. He noted that RIRs are not regulators, and engineers are not economists. It was argued that people need to think more broadly and understand the wider audience and contexts. It was noted that pragmatic choices need to be made.

    The presenter discussed the stages associated with coping with crises: Denial, panic, anger, blame shifting, bargaining, acceptance, recovery, revisionism. He concluded by suggesting we are currently at the panic stage.

    Questions and discussion

    • It was suggested that transition from IPv4 to IPv6 presents a great opportunity for the standards bodies to create new policies. It was asked if there were new policies for allocating IPv6 address spaces to authentic parties. The presenter noted that the allocation policy is currently based on abundant supply, and resources are distributed according to need. He suggested there is no point in artificially constraining supply when there is abundance, and that the current problem is the imposition of scarcity. The presenter said he is not sure if there is a need to impose artificial constraints on IPv6.
    • It was clarified that Internet didn't start business until 1993. It was noted that RIRs still have to justify their requirements to IANA.
    • It was asked if there should be a change to a completely unregulated economy, or something more gentle such as a regulated transfer model. It was noted that in an unregulated economy a lot of people lose. The presenter noted that he did not feel qualified to answer the question as he was coming purely from a network engineering background. It was suggested that this industry doesn't have all of the answers, and that there are questions that involve experts in other fields. The presenter advocated discussing markets with people who understand the issues and have had experience with similar commodities. The presenter stated that the questioners concerns are valid.
    • It was asked if this problem is worse for the AP region. It was also asked if APNIC would take care of ISPs or leave them to fend for themselves when in the case of an IPv4 shortage. The presenter responded that there is no difference between regions, and the issue must be globally orchestrated.
    • The presenter suggested that when RIRs run out of addresses they would no longer be allocators. It was noted having an organisation hosting a market is an entirely different issue to consider, and there are many players who need to be involved.
  3. IPv6 Transition & Operational Reality

    Randy Bush, IIJ


    The presenter described his presentation as 'reality therapy'. It was suggested that there would definitely be a transition to IPv6, and the issues to consider are 'when' and 'how'. It was argued that the marketing fantasy is not helping the industry to actually deploy IPv6 technology.

    The presenter noted that what should have happened is that IPv6 would have been deployed when IPv4 began to run out. However, IPv6 is not being deployed. Therefore the price of IPv4 will go up. The presenter suggested this situation has arisen because there is no long-term plan, no realistic cost, and no support on the front line.

    The presenter argued that IPv4 resource certification would make a hidden market open, visible, transparent and fair.

    The presenter noted that there are several myths associated with IPv6.

    The presenter argued that IPv6 transition is not easy, because it was designed with no serious thought about operational transition. It was asserted that IPv6 does not eliminate NATs because it is not compatible with IPv4.

    The presenter said another myth was that IPv6 would reduce routing load. He continued by saying that IPv6 space is not infinite, IPv6 security is not better, and IPv6 does not increase battery life. He also argued that deployment will not be incremental, and routers do not fully support IPv6. It was also asserted that IPv6 would not mean an end to static numbering. Ultimately, the presenter argued, IPv6 will not replace IPv4.

    The presenter argued that the key question is how to transition without losing anyone or anything, and the answer is to make deployment easier. This could be done by identifying and fixing transition problems, getting the IETF to fix protocol issues, and pushing vendors to provide support.

    The presenter argued that the industry should not pretend there are not problems. He also argued that giving away IPv6 space to promote it would create a mess; in addition, it is unnecessary, as IPv4 exhaustion would promote IPv6.

    The presenter noted that Japan in better shape with regard to IPv6 transition due to government financial support, and people with vision.

    The presenter concluded with a summary of the presentation's major points:

    • We must keep one Internet
    • The core needs to be dual stack during transition
    • There are five phases ranging from denial, through to IPv10 deployment
    • There will be NATs
    • We need NAT-PT
    • The IETF needs to face reality
    • NAT-PT and security issues need to be resolved
    • There is pressure on routing
    • Hacking is not a solution

    Questions and discussion

    • The Chair said he appreciated the comment that Japan is in good shape, and noted that the Japanese government has just initiated a program to decide what to do when IPv4 is exhausted.
  4. IPv6 - A positive "Approach" "Perspectives"

    Kusumba Sridhar, APNIC EC


    The presenter commenced by stating that his presentation was designed to market IPv6. He asserted that IP is an essential resource like oil, water, electricity, and spectrum. He observed in cases where those resources were depleted new technologies were developed.

    The presenter said moving to IPv6 provided technical benefits relating to routing, and also contributed to business continuity. The presenter suggested that IPv6 was a matter of technology, not policy. He also noted that it would not be a process of migration, but rather adoption and transition.

    It was noted that there is a public perception that there is a doomsday approaching where monitors will go blank. It was suggested that the future is much more positive. The presenter noted, however, that transitioning IPv6 would take a long time.

    The presenter contended that network operators would not need to make additional investments because IPv6 deployment would result in cost savings on security, NAT, maintenance and network management.

    The presenter encouraged organisations to establish IPv6 taskforces, and conduct an infrastructure assessment in order to begin planning for IPv6.

    Questions and discussion

    • It was argued that this type of presentation was dangerous because many assertions, such as improved security, were incorrect. The presenter thanked the speaker for his comments and noted that he was doing a marketing presentation, not a technical one.
    • It was noted that according to Geoff Huston's mathematical model, just before the exhaustion time, five /8s would be consumed in three months, which is about twice the current rate. It was asked if Geoff believed that level of demand would exist, and, if so, where the demand would be and for what type of technologies.
    • There was a discussion about the methods used to predict IPv4 consumption, such as demand extrapolation. However, Geoff Huston noted that in his projections he had not analysed the demand side; he had looked at history. However, he noted that some trends are interesting. He noted that IPv4 allocation in the North American region has been declining in the amount of IPv4 addresses being allocated relative to the other four RIRs. Of the other four, RIPE and APNIC each allocate approximately 35 to 40% each of all of the IPv4 addresses right now. LACNIC and AfriNIC, even though they're growing pretty rapidly, still don't have a large amount of activity. He suggested that the last few addresses were likely to be allocated by RIPE or APNIC, and that statistically, not taking into account a demand analysis, they are likely be deployed in Eastern Europe, the Indian subcontinent or China. It was suggested that the last few addresses would probably be used in a retail broadband deployment, possibly with VoIP components.
    • It was asked if APNIC plans to conduct a demand analysis. It was noted that JPNIC has started this kind of analysis and would like to collaborate with other organisations. Geoff Huston thanked the speaker for the offer of collaboration, noting that he has been as careful as possible to base his work on publicly available data.
    • There was discussion about the issues associated with introducing IPv6 into a deregulated communications industry. It was suggested that IPv6 has been a business failure in economic terms because organisations no longer want to invest in infrastructure. It was noted that the solution requires input from regulators, economists and people from many other areas.
    • It was suggested that special networks should not have to be built to deal with crises. It was noted that in order to make the transition to IPv6 happen, work needs to be done on systems available today. It was argued that part of the problem is the fact that currently this isn't happening.

Meeting closed: 11:25 am

Minuted by: Tina Bramley