Hi all,Geoff makes many points that are difficult to dispute -- e.g., that the industry has effectively missed its chance to execute a graceful exit from the current status quo (basically IPv4 + CIDR + the RIR system), that the status quo simply cannot support continued demand for IPv4, that preservation the registry *function* (i.e., not the mere appearance of a functioning registry) is absolutely essential to any/every scenario which assumes that the future builds on the accumulated efforts that have been invested in the Internet to date, that industry failure to preserve this function will necessarily and inevitably result in the emergence of one or more new, non- authoritative registry providers, and/or registries backed by sovereign authority.
I believe there is evidence of consensus on at least these three broad points -- no easy transition, insupportability of continued demand for IPv4 under "status quo" arrangements, absolute necessity to preserve the registry function... but perhaps this is something that merits an empirical test at the meeting in Christ Church (?).
If there is consensus on these three points, then I think it's worth carefully considering which course of actions are likely to be most "fitting" (in the evolutionary sense) for the continued growth and development of the Internet. Despite the near-exclusive focus to date on "resource transfers" as the only alternative to blind preservation of the status quo, there *are* other options. Given that, the question becomes: which is more important to preserve for the future of the Internet -- the supportability of continued demand for IPv4, or the number resource registry function?
Geoff and others have ably demonstrated that there are real and serious tensions between these two priorities, and has made a vigorous case for prioritizing the former over the latter -- or as he might say, accepting that the former is the only way to preserve the latter. I have argued that it is not the only way, and probably not the best way -- at least if there is one other point of consensus, namely that IPv6 *is* the chosen, community-supported successor to IPv4 as the "standard" mechanism for Internet attachment.
I do not expect to be in Christ Church at this point (although US Internet access conditions permitting, I will participate remotely), but I would very much like to see this question formally posed during the upcoming Policy SIG.
Here's why I think this question is important. If IPv6 is the recognized, community-selected successor to IPv4, then that may provide us with a framework for distinguishing what kinds of continuing IPv4 demand are supportable (or at least, merit ongoing community efforts to support), and which are not supportable -- or at least, merit different consideration by the community. For example, if IPv6 enjoys this consensus-supported status, then continuing demand for some quantity of IPv4 by (otherwise IPv6-based) new entrants would be consistent with that IPv6-centric view of the future, at least as long as new entrants absolutely need backward compatibility with the universe of IPv4-attached network resources. However, if IPv6 does indeed represent the future, then the same might not be true for other expected demand drivers for IPv4. For example, does the continued acquisition of IPv4 as a means for established, IPv4-based operators to postpone incorporation of some IPv6 into the service platforms merit active community support? Does it merit the same level of community interest as the case of new entrants, who will have no possibility of even establishing administratively independent networks without some small quantity of IPv4? In a world where IPv6 is the recognized, accepted, presumably inevitable future for all new and growing ISPs in the future -- and community support for mechanisms to postpone that reality could entail serious, permanent tradeoffs, then it's not clear (to me at least) that they do. Should acquisition of IPv4 by enterprises with no particular interest in public or private networking, and no intention to attach new resources to the Internet -- or worse yet, an active intention to deter the efforts of others to attach new internet resources -- merit any consideration at all by the community?
If it turns out that there is a consensus-supported "hierarchy of demands" for IPv4, and that hierarchy does not give equal weight to the two "difficult" cases -- i.e., IPv4 demand to enable basic administrative independence in the future, vs. IPv4 demand to postpone or avoid the use of IPv6 altogether -- then this observation could be used to develop alternative mechanisms for IPv4 recirculation that might not necessitate all of the certain tradeoffs and uncertain risks associated with prop-050, such as those identified by Izumi and the JPNIC community.
Moreover, such a hierarchy of demands would have important implications for that other demand driver that no one ever talks about on the policy lists -- i.e., the demands of surplus IPv4 resource holders to monetize that status, and profit from the sale of IPv4. Following the same chain of reasoning outlined above (granted, all speculative unless/until consensus support is demonstrated), the community might have a strong interest in supporting those IPv4 sales aspirations if they enable future network operators to enjoy the same range of opportunities enjoyed by community members to date (e.g., to be a multihomed, "provider independent" network or a competitor in the IP transit market). However those interests, and any tradeoffs that they required, might be substantially different to the extent that they enable existing IPv4-based operators to temporarily or permanently avoid their own IPv6 transitions. Again, if such a hierarchy exists, the community might even have an active interest in discouraging the latter, since every IPv4 sales transaction that contributes to delayed IPv6 migration -- especially by the largest operators with the most popular content and services -- could help to perpetuate the other kind of demand (i.e., the need for new entrants to have IPv4 for backward compatibility), and may even contribute to the gradual erosion of support for IPv6 itself as the foundation for ongoing Internet growth and development.
I'm going to stop here so that this message doesn't become (even more) fatally long and speculative. However, I would like to point out that prop-062-v001 (Use of final /8) may be viewed as one method of prioritizing the first kind of demand. There *are* other conceivable methods for accommodating the other sources of demand, and for creating incentives for increasing the efficiency of IPv4 utilization, which would likely entail fewer/lesser risks like those enumerated by Izumi, and also lesser risk of impeding or aborting the gradual shift to IPv6. If the consensus I've speculated about above is real, then I'll look forward to helping to identify some of those alternatives in the coming days.
regards all, TomNOTE: Some will probably claim that framing the question this way implies some kind of conflict or moral distinction between large and small operators, or "legacy" operators, incumbents, and new entrants, etc.... so to preempt any misunderstandings or mischaracterizations, I will again highlight the fact that I am asking for a public evaluation of consensus -- specifically for the unqualified proposition that IPv6 is the the chosen, community-supported successor to IPv4 as the "standard" mechanism for Internet attachment. If there is consensus on this point, then any conflict that exists is not between different classes of network operators, but rather between the short-term and long-term interests and incentives that *every* individual network operator must reconcile in the process of navigating the unavoidably challenging transition ahead.
On Aug 23, 2008, at 3:03 AM, Geoff Huston wrote:
Hi, Thank you for forwarding these comments and observations from the JPNIC community.Hi all, A few last minutes comment before the meeting... ### WARNING: it's long ###WARNING : its even longer now! :-)We undertand this proposal restricts APNIC's role to updating DB registrations. However, we believe implications and issues outside DB update must be also considered before implementing this policy.There is a view these days that where we are heading with IPv4 is not a pretty place for many reasons, and thats a view that I personally share. But to avoid that situation where the network's continued expansion is dominated by considerations related to unallocated address pool exhaustion, perhaps this industry should've made its moves to adopt Ipv6 quite some years ago, rather than leaving things to well after the last second to figure out how to avoid some of these issues. But thats now a lost opportunity and we are facing some pretty tough issues as an industry. And central to those issues are the management of addresses and, given that the allocation function is coming to an end, the ongoing management of the registry function. For me the most relevant aspect of IPv4 policies these days is all about what actions are appropriate to preserve the integrity, accuracy and common utility of the address registry for the Internet as a whole. Without such an authoritative registry which is _accurate_ and _complete_ then you might as well give up and return to using smoke signals, because its possible that the Internet won't be useful any more! :-) If IP addresses are no longer assuredly unique and the association of an IP address with a certain end point is no longer universally acknowledged as a unique association then the network heads in a direction that leads to a collapse into address chaos. So how do you preserve the integrity and accuracy of the IPv4 registry when the current mechanisms for address supply vaporizes and yet demand for additional IPv4 addresses is still strong? That's theprimary motivation for me behind proposing the address transfer policy.Our understanding of this proposal is that it implies to allow tradingof IP address between LIRs which is likely to involve monetary exchange.The policy explicitly does not cast APNIC into any role regarding the operation of monetary exchange, nor any role regarding the pricing addresses, nor the mechanisms relating to the way in which the exchange may have been arranged. Some would argue that a registry, or title office, has no role in facilitating a marketplace, and should remain strictly neutral. The transfer policy advocates placing APNIC in such a "title office" role in so far as the proposal does not place APNIC as a facilitator nor even an advocate of markets, nor does it place APNIC in a position of having to deny the existence of markets for addresses, if or when they emerge. The policy simply proposes to allow APNIC to recognise and register a transfer of addresses if the parties involved meet a number of nominated criteria. How the parties arrived at the position of requesting a transfer is explicitly not cast as being APNIC's business in the scope of this policy proposal.We've listed possible issues below: 1. Change in the nature of IP address/IP address distribution - IP adress may be considered an "asset" or "property" rather than a resourceI'll respond to this comment here, but with the caveat that this response is not really relevant to the policy proposal per se, and the comments are, as per the disclaimer below, strictly comments that are my opinions, and not necessarily positions that relate to any other party or any organization, including my employer. If you strip off all the layers and look at this in economic terms, what allowed addresses to be considered as valueless commodities was their abundance. If you had a need, the need was met from the allocation pools at effectively nominal cost. If you did not have a need then your valuation of the price of addresses was effectively zero, and if you did have a need then the allocation registries could meet that need at a minimal transaction cost. So the formation of any market in addresses was stymied simply because there were no buyers willing to pay because all such needs could be met from the existing distribution system. This lead to an observation that *within the context of that particular distribution framework* addresses had no intrinsic value and consequently could be considered as something other than an "asset" or "property" simply because they had no explicit monetary value within this form of distribution system. But such an observation says nothing about the intrinsic nature of addresses and says more about the properties of a distribution system that exploited abundance to suppress any organized formation of secondary markets. So when that distribution framework grinds to a halt due to supply exhaustion do the properties of a lack of a monetary value for an address still hold? In a conventional view of trade, economics and markets this would not be the case. Where a good is in demand and the demand exceeds supply then the relative scarcity is expressed as a pricing function. Now one view would be that this pricing function generates a consequent interpretation of the good as an asset simply because there is this value that is placed on the good because of the shift in the demand and supply situation. In other words the nature of a good, including addresses, is dependent on the context and the demand / supply position, the distribution framework and the utility value and the content of the utility function. Which is a long way of saying that what will change the nature of IPv4 addresses in terms of their view as an asset or property is the situation of premature exhaustion where demand will outlive supply, and this is not intrinsically an outcome of this transfer policy per se. i.e. this shift in the generally viewed nature of an IPv4 address will shift irrespective of this policy proposal. And if this shift in the value of IPv4 addresses presents an intractible problem, then perhaps the time when this situation could've been avoided has already come and gone and what we are left with is to accommodate what needs to be done as effectively as we can.- This proposal implies that there are no longer needs to demonstrate the needs in receiving address space. This leads to inconsistency that those who receive the space via APNIC must demonstrate the needs and those who receive it via transfer don't need to do so. It may result in those who do not necessarily need address space are eligible to receive it for financial reasons (We understand too strict a criteria would lead to black market, but perhaps there should be checks in some form to prevent hoarding)Most of this proposal is aimed at the post 2010 situation when the current distribution framework for Ipv4 addresses is no longer viable because the existing framework would've come to a halt. This comment therefore refers to the period between when this policy isimplemented and the point when the address pool is exhausted for APNIC.Now its my personal guess that this period will be short. Maybe I'm taking a pessimistic view of the speed of the policy development process these days, but proposals such as this one do take time, and it may well be that the above situation will not eventuate simply because exhaustion would've occurred prior to this policy being adopted.2. Social implications as a result of allowing transfer - IP address holdings may be subject to taxation for LIRs as well as APNIC. Impact of such change is not properly analyzedSee the above comments - once addresses are scarce and when demand outstrips supply then a pricing function will emerge. Whether this emerges in the context of the RIRs or in the context of alternative registry systems is yet to be seen, but the basic dynamics of the situation is unaltered. Irrespective of the RIRs' position this pricing function will have implications for address holders, In other words this comment appears to have traction *irrespective of the fate of this transfer policy* It is a matter of some concern given the potential range of price outcomes, but for any enterprise to suddenly have an incredibly valuable asset appear on their financial books is a source of some concern for a number of reasons, including, but not limited to taxation. So I would argue that address pools will have a significant asset value post IPv4 address exhaustion and irrespective of whether such addresses move between parties or not in the context of markets may attract taxes and changes, depending on the local rules and regulations concerning the holding of assets.- APNIC may get caught in legal disputes over adress holdings and transfers and restricting APNIC's role to DB update in the policy may not be effective enough as a protectionAgain, once the current IPv4 distribution framework grinds to a halt in the coming months, folk will search around for "unused" space and attempt to use it, and disputes will increase and recourse to legal remedy will always be available. Again, this policy is entirely neutral about this - the likliehood of escalation in the number of such disputes is independent of the adoption or otherwise of this transfer policy, in my view.We also had some discusssion in JP-OPM and major comments received were as follows: - Those who trade in black markets always chose the easiest way - therefore, even if this policy is implemented, they will likely to move address space between themselves without going through the official procedureI am not so pessimistic as this, and note that this (the formation of black markets) would be a terrible outcome for the Internet. The association of a holder with an address is not useful if it is kept as a private secret - for the network to work in any meaningful and useful way this association MUST be public and MUST be universally acknowledged. Black markets, if one defines such markets as occluded or hidden transactions where the parties and the outcomes are not visible to all others are useless - useless to the participants and useless to the Internet! The more likely scenario is that needs will be expressed in other ways. If parties wish to effect movement of addresses in order to meet continuing demands for addresses then if the RIRs are not able to record the outcome of such transactions then its more likely that other public registries will appear. Given that such registries would be holding current rather than historical information about the disposition of addresses one possible (probable) scenario is that everyone else would use those sources of information as the authoritative source of information about addresses. At that point the RIRs would revert to the status of an historical institution with no further relevance in the IPv4 world. In other words, denial of a shift in requirements for the registry function may well be a move which simply ensures that the registry function moves elsewhere where requirements are accommodated. But such an orderly transition is unlikely in this situation. A more likely scenario is a number of such alternate registries appearing each with differing information and differing modes of operation - at this point the ability to determine that any address is unique and the identity of the current holder of the address vaporises and the Internet itself then suffers potentially irreparable damage. The remedies to such a scenario are typically political and typically work "consensus" in the direction of "equal distaste expressed by every party". Again its not a pleasant scenario when compared with what we've been able to achieve in terms of cost, efficiency and efficacy in the RIR system.- Some form of demonstration of needs should be required to prevent address hoardingIts not clear to me that a title office should also act as a market regulator. Its not clear to me that APNIC should undertake such a role. That's not toi say that there should or should not be some form of restraint on the actions of individual actors that provides appropirate disincentive for hoarding, but it is saying that its probably asking a bit much for a title office to also undertake the role of a market regulator. To the extent that hoarding is a natural expression of the expectation of price escalation then surely the issue here is that at some point the alternative, that of more visible and rapid moves to quit IPv4 and use IPv6, becomes economically so much more attractive that the hoarder is left with a valueless pile of integers. So its unclear to me that address hoarding would be an attractive proposition financially given that there is substitutability, and this substitutability becomes more attractive the higher the price moves in IPv4.- simply register them as assignments rather than treat it as transfer of allocations. this will allow to achieve the goal without fundamental changes in the policy (issues on this methods were also discussed)This appears to be a continuation of the previous comment, that APNIC should act as a market regulator and constrain who can be a market actor and the nature of each and every transaction. Its an interesting proposition, but it certainty has some substantive issues, including issues of authority, credibility, competence, effectiveness, acceptability, and substitutability to name a few. regards, Geoff Disclaimer: Most of you have figured it out by now, but for those who haven't the above is entirely a product of my own delusions and definitely noone else's.* sig-policy: APNIC SIG on resource management policy *_______________________________________________ sig-policy mailing list sig-policy at lists dot apnic dot net http://mailman.apnic.net/mailman/listinfo/sig-policy