On Jul 21, 2009, at 5:04 AM, Eliot Lear wrote:
Hi Tom,
Thanks for the response. I actually agree -- probably should have written that competition was never more than one of many reasons that someone (anyone) *might have* supported those earlier transitions. That said, if you look back at my original response you'll note that I was actually arguing *against* transition rationales based narrowly on "competition" questions. The alternative that I proposed was transition to preserve the "openness of the internet addressing and routing system," especially to new entrants.
Sure! I particularly like the idea, for instance, of using LISP or other similar technology to enable end systems to move within the topology without having to pollute the global BGP table. But I do feel a Hal Varien moment coming on: are incentives aligned for the right things to happen?
Hi Eliot, [note well: what follows represents personal opinion** only] Thanks for (re) injecting these questions. On the "Hal question," I interpret the current state of affairs as as a weak indicator that incentives are on balance (i.e., far from perfectly) aligned. That said, I would never have allowed myself to get stuck in this rather uncomfortable position if I believed that that balance of incentives were actually aligned toward a sustainable outcome.
If not, can they be so aligned? Our earlier work pointed out that there is a limit to the efficacy of RIR policies. So what tools do we really have?
Incentive alignment might qualify as a "normal" engineering challenge under other circumstances. In this case the task is abnormally complicated by (a) the apparent absence of strong internal consensus** or other means of identifying the common goal, and also by (b) the absence of half of the engineering tool box (i.e., all carrots, no sticks -- and while our carrots may be very nutritious, no one has ever accused them of being irresistibly delicious). Ideally, our unique community decision making (aka consensus-seeking) mechanism would provide a means for overcoming these challenges, by providing a forum where contending views on both policy means and policy ends are exposed to cross-examination. However, in this case, I have a feeling** that only one side of the debate is actually being presented openly. If we were to take the decreasing frequency of open, public objections to IPv6 transition/integration as a de facto indication of support for such a move (which seems to be a common feature of the consensus-seeking process, at least toward the end stages of policy development), my guess is that we'd probably expect the transition to be much further long, and to be happening at a much faster, and at a more rapidly accelerating pace than available evidence would seem to suggest. So, am I just fundamentally wrong on some point of fact or interpretation, or it is possible that some unknown share of the apparently widespread non-objection (i.e., the proverbial "dogs that don't bark") actually reflect some (equally unknown) number of unannounced decisions in favor of non-transition? I don't even pretend to know the answer. But my doubts have prompted me to take what seems to be the most institutionally appropriate action -- to capitalize on the one good (and perhaps unique) "tool" that we do possess, i.e., to use our community deliberative mechanisms to try to present a positive case *for* a timely and orderly transition. Though I may be lacking in the skills required to wield them effectively, for now they remain the best (maybe only) tools that I have at my disposal.
Again, returning to the LISP example, it's a technology advance that is a potential Over-The-Top play that could align interests (not that I've done the analysis, but that's my off-the-cuff thought, something to be wary about).
I think that this possibility merits much more thorough and systematic consideration. My (quite possibly under-informed) impression is that widespread adoption of LISP might improve matters substantially in several ways, e.g., (1) by providing the majority of origin-only/self- service routing providers with an opportunity to economize by vacating the DFZ, (2) by reducing the likely ongoing demand for DFZ participation (and al that that implies in terms of RIB/FIB scalability) by permanently eliminating such participation as an involuntary requirement for self-service routing providers, and (3) by reducing the impediments to renumbering, which hypothetically might contribute to the absolute volume of transferable IPv4 resources available to aspiring new entrants in the post-IPv4 exhaustion era. If this is not too far off, then (assuming LISP succeeds on all of the above counts) I might still be curious about the (extent of the) dynamic aspects of LISP's scalability advantages (e.g., how much of the expected DFZ gains derive from the one-time payoff arising from the voluntarily exit of non-transit providers, and how much/ for how long would the remaining uncommitted DNZ capacity be likely to remain "open" to future new entrants?). Also, even if LISP ultimately succeeds in making 99% of current IPv4 assignments technically unnecessary, I would not expect that fact to automatically/inevitably or dramatically alter the "scarcity premium" that IPv4 will likely command in any future in which IPv4 remains non- substitutable for any critical industry functions or roles, and protocol resource numbers are (re)distributed exclusively through market mechanisms. Recall that back in the late 1990s, very few people actually new that 99+% of even the initial lit capacity across all of the segments of all of the recently deployed optical systems was completely idle. Prices didn't start to take their substantial (!) downward turn, and corresponding "productive capacity utilization" rates didn't start to really explode until the actual facts surrounding those supply condition were inadvertently revealed a few years later, in the course of a few unanticipated distressed asset sales, legal investigations, etc. In general I think that the assumption that there is any automatic or inevitable or guaranteed relationship of any meaningful and reliable kind between supply and demand has been largely discredited by history. Markets often work quite well, and (perhaps) well-designed and competently tended markets tend to work even more-better-more-often (?). But I think that the era when anyone can assume that the mere coincidence of some supply and some associated demand automatically guarantees the emergence of the best of possible worlds over any timeframe relevant to human beings has now passed into the history books. More specifically, in this context I think that the tendency to assume that market mechanisms would infallibly convert the fact/existence/ supply of additional loose (de-assignable, transferable, etc.) IPv4 into the enduring condition of greater openness to aspiring new entrants is based on the widespread (mis)perception of individual IPv4 addresses and prefixes as "things" (assets, commodities, etc.), when in fact they actually represent something more like discrete instances of a "generalized privilege" or "license" (as in "creative license" even more than "driver's license"). If/when future technology advances provide a /32 owner the same kind of "generalized privilege" that today is limited to those holding a /24 or more, then I would fully expect the "privilege adjusted" cost to remain constant, if not appreciate. As long as IPv4 remains non-substitutable for any critical and/or commercially attractive aspect of content or routing service provision, that "privilege adjusted" price is only likely to continue appreciating ad infinitum.
I think that the majority of reasons for supporting the previous transitions (i.e., previous impositions of new technical requirements on incumbent network operators), as well as for supporting IPv6, would fit neatly under this rubric. And while your own description of the past circumstances didn't use those specific terms, I believe that the phenomena of "proliferation of class Cs being allocated" and "class Bs running out" that you referred to were both driven in part by demand from initial allocation seekers, a.k.a. "new entrants" -- i.e., we agree on this point.
Of course, during any given duration the majority of newly allocated/assigned IP addresses are probably going to "incumbents" (subsequent allocation seekers), but that's consistent with patterns of growth in every other industry, i.e., it's basically a demographic of statistical artifact of cumulative growth processes (granted, one that can be influenced by various policies). The important distinction, at least until now, is that the growth/IP addressing demand of incumbents was not incompatible with the continued openness of the addressing and routing system to non-incumbents/new entrants.
And of course, I wouldn't have expected these sorts of arguments to figure prominently in IETF et al. meetings and discussion back in the 1980s-1990s. In the main, we're not a bunch of economists and policy makers by profession -- and even those of us who are generally know better than to phrase our technology and policy advocacy in such alien and unsympathetic terms... most of us anyway.
Well but as you know there were some discussions along these lines in the CIDRD/BGPD working groups. However, there are limits as to how far those discussions can reasonably go in that forum.
Of course you are right. But as noted above, I know of no better options at the moment.
One thing I would draw your attention to was a paper in this year's WEIS conference by Richard Clayton that contains an economic analysis of SHIM6. His parting question from his presentation was whether RFCs should contain Economics Considerations sections.
This sounds very interesting, although I'm not sure that it would be easy (or even possible) to subject an Economics Considerations section to IETF-style quality and relevance-preserving mechanisms -- and I'm even more uncertain whether an Economics Considerations section that lacked such vetting processes would be likely to net out positive in terms of fog lifted, questions answered, etc. As the ongoing TV-MM dialogue should make abundantly clear, achieving consensus on economic fundamentals, or the validity of different methodologies, or the relative value of different knowledge bases or experiential inputs, etc. makes IETF or RIR decision making look like the ultimate paragons of both decorum and efficiency. That said, I do think it would be well worth exploring.
However, even if "openness of the internet addressing and routing system" was never the most popular/resonant rallying cry for specific technology changes, that doesn't mean that this general idea wasn't an important motivator or an influential part of the unspoken context. And even if that sounds implausible too, it still doesn't mean that the *fact* that such openness was preserved, intentionally or otherwise, didn't play an important role in the success of the system to date. Although the verdict of history is never truly final, to me industry openness looks a lot like a universal correlate (i.e., a necessary if not sufficient cause) behind the durability of industry self-governance. If that turns out to be true, then there could be a lot more at stake in the current transition tussle than just the fate of the IP address registries.
I don't think anyone can deny that openness was an important motivator for Internet technologies. We merely need to look at all of the OTHER protocols that have gone by the wayside to see that it is so. But I think what we saw was that Opennness was something to be capitalized upon by one group of vendors, to entice customers, and to disrupt certain market players. It certainly was not regulated.
I agree entirely -- which is why I've been trying to articulate an *evolutionary* understanding of the relationship so far between industry self-governance and rapid Internet adoption, growth, and development. I have never claimed, nor do I subscribe to the view that an ongoing, conscious, active pursuit of openness as a top priority by industry direct stakeholders is responsible for the Internet's great success. Rather, I'm suggesting that the generally welcome but largely accidental/unintentional preservation of industry openness throughout previous system-wide technology transitions and continuous rapid growth probably represents a major (i.e., almost certainly necessary if not entirely sufficient) reason why this world-spanning critical infrastructure continues to be entrusted to (mostly civilian, mostly private sector) full-time technologists who occasionally take a few days off to serve as volunteer lay economists and policy makers. IMO**, those present at various critical moments accidentally succeeded -- perhaps with the occasional assistance of an unusual conjunction of "selective incentives and joint products" [[1]] -- in creating a very durable, relatively adaptable, effectively self- maintaining industry coordination mechanism. So far, those who came along thereafter seem to have done a reasonably good job of keeping it running, and headed in what seems to be a generally benign direction. (Needles to say, there's always room for improvement) I don't think that anyone has the power to recreate the kind of circumstances that give rise to this incredibly fortunate albeit accidental/unintentional evolutionary outcome. But one could argue that, for every system/organism or whatever that survives long enough, there inevitably comes a time when self-awareness coupled with the freedom to act replaces (or at least substantially complements) the blind mechanism of evolution in determining what happens next. Becoming aware (or being reminded) of the fact of evolution and its possible impacts will not relieve anyone of the important choices that will be made, one way or another, in the very near future. But I do believe that it might contribute (even if very marginally and indirectly) to better decisions.
And indeed if we look at security and authentication technology, the desire to play King of the Mountain has IMHO long stalled improvements in consumer authentication.
Even though I'm not clear on the specific context, your frustration comes through loud and clear. I think that the quest to develop a perfectly open, self-encapsulated/externally unencumbered, and self- maintaining identification/reputation technology has been with us for a long time. Maybe when (if) we come up with one, it won't create any new complications that are equivalent to the current sources of frustration with external authority-based authentication systems (which have also been with us forever). I can already think of a couple of plausible candidates, but then that's my job -- even so, I always love surprises ;-) Regards, TV [personal opinion only] [[1]] The "joint products" model is an explanatory device that economists have used in some cases to describe how public goods like "industry openness" are sometimes produced more or less accidentally. It has been used, for example, to explain some historical episodes in which bankers in decentralized monetary systems -- basically countries where individual privately owned banks were at liberty to print their own currencies as they saw fit, which tended to be quite extremely volatile -- voluntarily adopted a standardized, national currency and "industry self-governance" model with some binding rules (e.g., capital reserve requirements, money printing restrictions) that implicitly limited their commercial freedom. The basic idea is that in some cases, the individual industry players themselves are confronted with some new opportunity to realize a substantial private benefit, but only if they cooperate in some specific way (e.g., in this case the adoption of a single national currency) that has the collateral effect producing some non-private "meta-good," like a national economy where individuals and businesses don't have to worry very much about whether their money will just stop working (i.e., their private/ nonstandard banknotes cease to be accepted by merchants), or their bank-deposited savings will suddenly disappear without warning.