Hi Geoff, Thanks for the response. This is all very interesting! On Jun 13, 2008, at 11:09 PM, Geoff Huston wrote:
Tom Vest wrote:
The only question -- and the one that people seem to be unwilling to engage -- is whether this particular prohibition was always inherently silly or rather important in its own right, whether the infraction to be legalized was/is trivial or truly dangerous.
This strikes me as a generalization arising from a false premise - namely that the external environment has not changed in any manner, and the change is purely a change in the manner in which the address distribution function occurs. Of course that is simple not the case here, which becomes apparent in looking at the question you then posed:
ftp://ftp.ietf.org/ietf-online-proceedings/95apr/area.and.wg.reports/ops/cidrd/cidrd.rekhter.slides.ps Was Yakov always wrong? Has something changed to make him less wrong today?
Was Yakov always wrong? no
Is he less "wrong" today - dunno - he was right back then, but the environment HAS changed, so the "less wrong" question is rather irrelevant isn't it?
As most folk on this planet are aware (and elsewhere too these days, considering the level of air play of this particular story :-) ), the environmental shift from the perception of abundance in IPv4 addresses at the time to a current perception of scarcity and looming exhaustion is a massive change in our environment. This percpetion is now driving industry behaviours, and I'd claim that the steady increasing address allocation rates over the past 2 - 3 years are not unconnected with this shift in perception.
The observations that apparently drive much of today's consideration are:
- this industry is just too large, too diverse and too bound to low margin commodity operations for much of it's Internet activities to drive (and fund) the completion a comprehensive transition to IPv6 within the timeframe available as defined by the remaining IPv4 unallocated address pools.
This is an interesting observation. How much would you say is the additional marginal capital cost of migrating to IPv6, for a network operator that has purchased network gear in the last 2-3 years? Presumably anything older that that will be coming up for refresh anyway, and anything newer than that would be pretty good to go, wouldn't it? Do you think that the costs scale roughly in line with the stakes involved -- i.e., that large operators with large capital stocks and large market shares will face greater costs than smaller operators, or do you think that the gross costs and benefits are somehow misaligned? If they are aligned, then I'm not sure why the size or the diversity of the Internet should have any bearing at all on this. Doesn't competition help to reveal when margins are really "too low" -- in fact, isn't competition the only way to know what margins are, or what they should be? And if there is no competition, calculation of margins are pretty arbitrary, aren't they?
The demand for addresses, as defined by the prevailing needs-based address distribution policies and the associated industry demand levels, appears to extend beyond the anticipated point of exhaustion of the current supply system.
I think everyone would agree that the demand for *logical attachment to the Internet*, as currently/contingently illuminated by the IPv4 run rate, clearly shows that demand is going to continue beyond the point of IPv4 free pool exhaustion...
- this implies that there will be a continuing need for IPv4 addresses after the RIR pool has exhausted within this industry.
Okay, so the demand for attachment will continue unabated, and technically that demand can only be satisfied, directly or indirectly, with IPv4 resources -- because IPv6 is not backwards compatible "on the wire", and because all of the edge resources of interest to current and future end users are now attached via the IPv4 resources of incumbent operators.
- It is also observed that there is considerable diversity in the current value of "use" of addresses today, and it is likely that this broad diversity of exploitative value of addresses will persist.
That's an interesting way to phrase things! We are confronting a zero sum game, in which "low value" resources that cannot justify the value of their attachment will go under so that more valuable resources may be attached. Actually, since that value calculus will be expressed in terms of willingness to pay, perhaps we should restate in clearer terms: big money players will buy out small money players, until an equilibrium has been reached where no player is so small that it can be bought out by another.
It appears that in this case once the existing supply mechanisms have been exhausted, then the supply mechanisms to meet the continued demand will come from existing holders of address space. Its also pretty clear that such supply mechanisms will not meet every last possible demand, so a regime of scarcity-based redistribution will eventuate to equilibriate supply and demand. In such scarcity-based redistribution mechanisms, those applications that place a higher exploitative value on IPv4 addresses would be capable of making a case to transfer addresses from an existing address holder who has a lower value associated with the addresses.
So far I think I summarized pretty accurately... if somewhat less clinically.
This equilibriation is conventionally based on establishing a valuation point whereby the residual demand levels can be met in that those demands where the exploitative value is still higher that the valuation point will be satified, and those who have placed a lower valuation on meeting their demand will seek substitutes.
Ahh okay, so what are the available substitutes, and what is the scope of their substitutability? Would you say that an aspiring new entrant that wishes to multihome sometime in the eternal post-free pool age that is about to begin -- and actually provide normal ("full" interdomain) Internet access to customers -- would be able to go into business with IPv6 alone, or RFC 1918 space alone, without a single IPv4 address? If not -- if they will need at least that a couple of /32s to attach to other networks -- then in effect you have conceded that there is no substitute, at least for future network operators that aspire to enjoy a level of resilience, flexibility, and "autonomy" equivalent to that which has been available to every network operator to date. What would have to change in order for the possibility of "true, full" substitution to be restored, and who would have to change it? Won't that day have to wait until a preponderance of Internet resources -- users, content, services, etc. -- are "made accessible", either by being renumbered into IPv6 themselves, or via some other kind of mediated service, either of which would have to be undertaken by incumbent IPv4 operator/seller -- i.e., the same institutions that have been incentivized by the transfer market to capitalize on the value of IPv4 for as long as possible? Now why would a service provider in a hotly contested market, especially one that is otherwise "too bound to low margin commodity operations" *ever* willingly give up that market lever and revenue opportunity? Even if one attempted to, their high margin IPv4 resources would just get snapped up by a more aggressive competitor, and nothing would change. Please point out the flaws in my logic, because this sounds to me like a perfect, inescapable trap -- one that closes the industry to all but superficial/subordinate entry permanently, and guarantees that the transition to non-dependence on IPv4 could take forever, literally.
Now all of this would be irrelevant if we were still in an environment of abundance - all those who had a case for addresses, whatever the exploitative value each party places on the addresses, could be met, in which case you get back to the proposition that the only residual 'value' of a particular address over any other is the relative ease with which it is routed - which is what Yakov's slide pack is effectively saying.
I think I understand your point, but if I do this is just a restatement of an old economics joke. Econ prof and student walking across the campus, both look down and see a $20 bill as they step on/over it. Prof never breaks stride, student asks "But Professor, why didn't you pick up the money?" Professor says, "I didn't bother because it couldn't have been there -- if it was, somebody else would have already picked it up!" So, the value of routes will be attested to by the fact of their being routed, and the nonvalue of nonroutes will be equally self-evident...
But this is all basic theory of markets, isn't it Tom. I don't I've said a single thing about market valuation, exploitative values and substitution that isn't in any standard text on the theory of markets!
I should have read more of standard texts I guess. However, I never found that they had much relevance at all in this sector. In this sector, however, I have seen countless examples of critical, non-substitutable, bottleneck inputs being strategically rationed by incumbents, with the sublimely rational goal of precluding any/all competition that did not guarantee them the lion's share of any increased revenue or market growth that anyone realized as a result. An equal share of a bigger pie is never enough -- that's just economic rationality at work.
Is routing so fragile that addresses cannot move at all? I think not.
Is IPv6 so busted and NATs just so unworkable that any form of substitution is completely untenable? I think not.
I eagerly await clarification in this substitutability question -- because if I'm wrong, perhaps things are not so dire.
Are there addresses out there that are unused, or used in contexts where subtitution may take place? I believe so.
Lately I hear this as justification for the assumption that all "underutilized" IPv4 will be consolidated in 1-2 years post-runout. So let's just stipulate this, and focus on what happens after that.
So I'm left with a question or two back to Tom via this list: Why do you think that in this case the conventional mechanisms that we've use explicitly or implicitly to equilibriate demand, supply and substitution within human activity sectors for, oh, a millenium or five, in times of scarcity of supply and competing demands have to be suspended and some magic substituted in its place just for IPv4 addresses?
Actually, there is an important class of finite resources for which the "conventional mechanisms" were never applied, precisely because the resources in question played an integral role in facilitating all of the other exchanges in the economy, helping to keep them "efficient" with low transaction costs, etc. These resources were also generally held in escrow by a central authority, which was the only authorized recipient of freshly discovered stocks of the resource, and also the only authorized top-level source from which others could secure this resource. Like all growing systems, these eventually hit a ceiling with their finite critical resource also, and had to transition to a different quantity-unconstrained substitute -- episodes which may provide some useful insights for our own predicament. I can tell the full story later if there is interest, but suffice it to say that there are other relevant precedents and mechanisms.
And the second question is: Equally I'm really unclear what you see as the alternative framework to use here to address the situation. Do you believe that there are credible alternative approaches that do not involve the movement of IPv4 addresses between entities beside the "lets do a complete IPv6 transition in the next 24 months across the entire globe" approach?
Since that's not a credible alternative either, let's hope so! Actually I do have an idea, but it would require that the tempo and discomfort of the transition to be distributed in direct rather than inverse proportion to the stakes and capability of the parties involved. In other words, it would require that the largest operators (i.e., those most capable of transitioning, but also of postponing their own pain for as long as possible -- and also of directly causing the discomfort experienced by others to endure for as long as possible) to start their own (perhaps very very gradual) transition, sooner rather than later. Small and medium-sized operators too, the point being to get everyone on a transparent -- not necessarily equal or lockstep or aggressive, just transparently "confidence inspiring" -- timetable, at the end of which IPv4/IPv6 substitutability concerns are eliminated for all time. Note that this envisions nothing more than a modest acceleration of what the rosy scenarios predict will be inevitable in the 2-3 years post-runout, when all "loose" IPv4 will be spoken for -- and would preempt the sort of equally plausible but gloomier outcomes that some people (not just me) have described. Again, I could provide details later if there is interest, perhaps in another forum if that would be more appropriate... TV, speaking for self only