Re: [sig-policy] [Sig-policy] prop-062-v001: Use of final /8

  • To: David Woodgate <David.Woodgate at telstra dot net>
  • Subject: Re: [sig-policy] [Sig-policy] prop-062-v001: Use of final /8
  • From: Tom Vest <tvest at eyeconomics dot com>
  • Date: Tue, 29 Jul 2008 11:44:35 +0100
  • Cc: Philip Smith <pfs at cisco dot com>, 'APNIC Policy SIG List' <sig-policy at apnic dot net>
  • Delivered-to: sig-policy at mailman dot apnic dot net
  • In-reply-to: <200807290718.m6T7IeZ3071850 at burn dot telstra dot net>
  • List-archive: <>
  • List-help: <>
  • List-id: APNIC SIG on resource management policy <>
  • List-post: <>
  • List-subscribe: <>, <>
  • List-unsubscribe: <>, <>
  • References: <> <> <> <> <> <> <> <> <> <> <> <> <> <>
      Hi David,
      Thanks for the comments; additional responses inline.
      On Jul 29, 2008, at 8:18 AM, David Woodgate wrote:
      A permanently unused address does not contribute to the Internet community; any reservation of the remaining addresses must have a deliberate purpose, and have clear criteria for when and how they will be used.
      I agree entirely; in fact I would say that your assertion is perfectly consistent with the conservation policies and practices that have been applied to date, and which would continue under prop-062-v001.
      It may be true that a /8 is allocated within a few months at the current rate of consumption, but it still enables connection of 16 million people (or more with NAT) to the Internet. While I agree that 16 million is a very small proportion of the population in a region with 2 - 3 billion people, it is still 16 million people nevertheless.
      So, in a scenario where IPv6 is not available universally within 10 years for some reason (which I hope will not be the case), what would be valid reasons for making those 16 million people wait for connectivity? That is, what would justify reserving addresses for later use rather than allocating them in 2011?
      I think there's a serious misunderstanding here. IPv6 is "universally available" now, albeit at present universally inert -- unless one also possesses the means to translate and exchange packets with/across the globally ubiquitous IPv4 numbered hosts and other network resources. If some small quantity of IPv4 remains available to future new entrants -- *all* of whom will be otherwise exclusively IPv6-based -- then those entities will be able to internalize and directly administer their own IPv4/IPv6 translation requirements; in effect, they will possess much if not all of the same capacity "to be autonomous" that is now and has always been enjoyed by every network institution that has been eligible to secure PA or PI number resources from a regional registry. The reservation provides nothing more than that. However, that should be enough to assure that new entrants will not be so crippled by comparison to incumbent/legacy IPv4-based operators as to be relegated to a fundamentally different, arguably inferior, and permanently subordinate new industry niche.
      IMO, If you think the industry should remain open to new entrants -- or even if you don't care but would prefer to continue operating with substantial freedom from close antitrust oversight/monitoring/ intervention, then this is a minimum requirement.
      To my mind, the only reasonable cause to reserve addresses would be if there were an expectation that the ratio <connected services:IPv4 address> would greatly increase in coming years - that is, address reservation might be worthwhile if we could reasonably believe that instead of using one /8 to connect 16 million people in 2011, we could use it to connect 160 million in 2016, or 1.6 billion in 2021.
      Well, by securing the means to internalize their own v4/v6 translation services, the proposed reservation enables future new entrant content, service, access, et al. providers to enjoy the same level of operational independence -- arguably, the same chances of building sustainable new service businesses -- enjoyed by every other independent network operator to date.
      I suppose by implication that would satisfy your demand, and increase the <connected services:IPv4 address> density level -- but then so would allowing the market to become completely closed to new entrants and building all new/future services using NAT and IPv4 only. So I don't think that's a very good benchmark. What the proposed reservation will do, *uniquely*, is allow the <connected services:IPv4 + IPv6 address> growth trend to continue, with the smallest possible disruption given the otherwise unavoidable circumstances.
      (I don't know how such ratio growth might be achieved - it might be through something like facilitation of IPv6 deployment (say, 1 IPv4 address is used with 1000 IPv6 services), or through improved NATing, etc.)
      If no improvement in <service:address> ratio is expected, then it will only ever be 16 million people (or whatever current NATing allows) that can be connected, irrespective of when those connections happen.
      So, my problems with prop-062 from the perspective of the above points are: - It seems likely to incidentally rather than deliberately reserve addresses, without criteria for final usage;
      I cannot speak for the authors, but my own sense is that the criteria for usage is the same as it has always been for public IP addresses delegated by the RIRs: to facilitate the attachment and (at least potential) inter-operation of new users, content, services, etc. with the universe of other users, content, services, etc. that we call the Internet. To be perfectly honest, the reservation fulfills this purpose most clearly for new entrants -- i.e., "initial" allocation seekers -- but the decision to make it non-exclusively available on a one-time-only basis to all comers makes sense as a lightweight, pragmatic solution to potential complications like determining eligibility, etc.
      - Its premise seems to be based on equally sharing the pain of IPv4 exhaustion, rather than identifying how limiting distribution could provide true management across IPv4 exhaustion and IPv6 implementation, leading to improved Internet connectivity in the entire Asia-Pacific region across this time.
      No, its premise is that growth and dynamism and innovation are fostered by the decentralized management, parallelism, and competition that have been generally characteristic of the Internet to date -- and that those features are best preserved if the future, mixed IPv4+IPv6 addressing system continues to be characterized by as much of the uniformity/equivalence of technical use value that has been the norm for the IPv4 addressing system to date.
      If you wish to make a positive argument that "limiting distribution" of IPv4 would provide different, concrete benefits, then I for one would be happy to see the details.
      A suggestion for an alternative proposal might be one where future IPv4 allocations would only be made if it could be demonstrated by the requester that 1 IPv4 address would support connectivity for X hosts, where X is a number which may not be achievable with today's technology, but could be reasonably foreseen to be possible in 5 years' time or so (whether this be through IPv6-to-IPv4 translation, NAT or other means).
      Your phrasing is ambiguous. All things remaining equal, new entrants that emerge after the unallocated IPv4 pool is exhausted will be able to support intra-domain connectivity for an infinite number of hosts with IPv6 alone. All things remaining equal, post-IPv4 new entrants will be able to support inter-domain connectivity for exactly zero of those hosts -- unless they are able to secure means to perform IPv4/ IPv6 translation as necessary. This will remain true forever, or at least for as long as most of the "rest of the Internet" is not reachable by an IPv6-only network. The goal of prop-062-v001 (my own interpretation) is to assure that the distribution of this critical new translation functionality/requirement invariably follows the contours of the established number resource distribution hierarchy; those that would be "top-level" operators under the system now and in the past are not relegated to some degraded status in the future.
      Granted, that may be a far cry from what was once envisioned for IPv6, but at least it assures that the immediate future is no worse than the status quo.
      Again, if you could describe in more detail how the only alternative -- i.e., translation services becomes a permanent technical and commercial bottleneck, with incumbent IPv4-based operators the only possible suppliers -- would be preferable, then I think many subscribers to this list would be happy to respond, myself included.
      At 07:36 PM 28/07/2008, Tom Vest wrote:
      Given the stakes involved, and the eminently plausible outcome that
      some small quantity of IPv4 may continue to be both non-substitutable
      and non-optional for independent operations for a very long time to
      come -- much longer than ten years -- AND the fact that there will no
      way to remedy the situation if your assumptions about the time-to- tipping point turn out to be mistaken, wouldn't it be prudent to plan in a way that preserves as much flexibility as possible for an unknown
      future? As Randy noted, the run rate for sub-/8s is often measured in
      days... Is it really worth continuing on a course that everyone knows
      is destined to end for a few days more, at the price of giving up much
      freedom to adjust to unknown/changing circumstances in the future?
      No reductio ad absurdum reactions please -- the price of this policy
      as written is a few lost *days* of status quo allocation activity,
      nothing more.
      On Jul 28, 2008, at 8:31 AM, David Woodgate wrote:
      Thanks, Geoff - this is useful information for the discussion.
      It seems to confirm that the likelihood of getting to 8,000 LIRs in
      the next 10 years is very unlikely. (And I suspect that not all of
      the 4,403 LIRs to whom allocations have been made by APNIC would be
      active now.)
      At 03:13 PM 28/07/2008, Geoff Huston wrote:
      David Woodgate said the following on 22/7/08 14:10:
      My main problem is that prop-062 seems to risk locking up the
      majority of the last /8, and therefore does not share it at all,
      let alone in a fair and equitable fashion.
      I don't see how it is locking up the majority of the final /8.
      Would you please explain this.
      prop-062 allows for 16,000+ LIRs to each get a minimum /22
      allocation. As discussed in a previous email, it seems hard to
      justify even 4,000 LIRs over the next few years; I'd suggest that
      8,000 LIRs in the Asia-Pacific seems unlikely within 10 years. That
      would seem to leave up to 8,000-12,000 * /22s unclaimed for a long
      time. But - if I'm reading it correctly - prop-062 doesn't seem to
      suggest that anything else would be done with this unclaimed space, and therefore it won't be used during that time; that is, the space
      is "locked up" and unused.
      you make the claim that:
      "it seems hard to justify even 4,000 LIRs over the next few years;
      I'd suggest that 8,000 LIRs in the Asia-Pacific seems unlikely
      within 10 years
      Here's some historical data that may be useful in the context of
      this particular
      aspect of the discussion
      APNIC publish an "extended" version of the daily stats file
      ( latest")
      The last field in each row is a code for the end entity recipient of the address allocation or assignment, or approximately "LIR" in your
      Now there is some small uncertainty in the figures as at times the
      NIR code
      is used instead, but overall heres the Ipv4 allocation record for
      APNIC since
      2000, based on the numbers in that published file
      year new repeat cumulative count
      2000  94    432 2856
      2001  86    430 2942
      2002  83    339 3025
      2003 115    425 3140
      2004 120    570 3260
      2005 216    617 3476
      2006 253    786 3729
      2007 394    745 4123
      2008 280    429 4403
      i.e. in 2007 APNIC made 394 IPv4 address allocations to "new" LIRs
      and 745 allocations to LIRs who had already previously received an
      address allocation. Overall APNIC appears to have made allocations /
      assignments to 4,403 LIRs since its inception, and some 1,547 new
      LIRs have been recorded since 1 Jan 2000 (i.e the last 8.5 years)
      *              sig-policy:  APNIC SIG on resource management
      policy           *
      sig-policy mailing list
      sig-policy at lists dot apnic dot net