On Mar 8, 2010, at 4:54 PM, Brian E Carpenter wrote: > On 2010-03-09 00:21, Tom Vest wrote: >> On Mar 8, 2010, at 12:08 AM, Noel Chiappa wrote: >> >>>> From: Tony Li <tony...@tony.li> IPv4 is done. Over. >>>> Cooked. Fully toast. It will either enter a black market >>>> where we deaggregate and no proposal will help, or we >>>> shift to v6 and v4 is irrelevant. In either case, we're >>>> not in time to do anything significant for v4. >>> Probably needless to say, I do not agree with this. There >>> are _definitely_ more options than DeathI and DeathII. >>> >>> Noel >> >> There is also DeathIII, i.e., we shift to v6 (or something >> else), but through some series of steps that preserves IPv4 >> in place as a critical, non-substitutable input for >> provisioning routing services. If that happens, black >> market-driven deaggregation would not be inevitable. > > I think you're neglecting what is actually happening now, which > is a (semi) permanent shift to a dual stack routing network > with various forms of multiprotocol interworking, including > but by no means limited to v6/v4 packet level translation. > > That will allow the v4 network to survive for a long time, > but my personal guess is that it won't deaggregate beyond > about a million prefixes, because it will just be easier to > grow IPv6 after a certain point. Yes, there will be a fairly > messy grey or black IPv4 market and a lot of users suffering > from gangrene (double or triple NAT, or living on the wrong > side of NAT64). > > IPv4-only providers will indeed end up as dinosaurs. Funny, > really, that it's the incumbent carriers who seem to be > at the lowest risk of this. > > Brian
Far from neglecting, I've been closely following & actively encouraging IPv6 deployment, at least as much as my own circumstances permit. But while some recent movements in that direction are fairly encouraging, I personally think that it's premature to assume that they represent a "(semi) permanent shift" to anything, much less to the sort of dual stack routing environment that would enable IPv6-only networks to operate at anything less than a crippling/fatal disadvantage. In another, very closely related context, the correlates of transition that you identify -- messy, opaque, unpredictable input exchange markets, interworking mechanisms that impose various local and transitive disadvantages on some parties, while advantaging others -- have been identified with an historically recurring adaptation pattern called "Gresham's Law," which in this context would would translate to "bad interworking mechanisms drive out the good/better interworking alternatives." The mechanisms that have been historically associated with Gresham's Law are the exact same developments that we're anticipating today, e.g., 1. "Strategic rationing" of the more favorable interworking tokens by those who have them, for use only in their most critical interactions and exchanges, i.e., with (economic) peers and/or parties with superior bargaining power. As a result, whatever distinct advantages were provided under the previous interworking regime become increasingly rare and inaccessible, esp. to new entrants. 2. Deep aversion to using the less favorable interworking tokens by the majority of system participants, driven in part by a persistent, widespread, and very durable (b/c it's not strong informed by empirical counter/evidence) perception both that the other kind interworking token is not reliable now, and that it's future utility is even more uncertain. 3. On balance, the technical decomposability of key interworking inputs (in this case short < > long IPv4 prefixes, individual IPv4 addresses < > TCP ports) actually encourages as much or more aggregation/consolidation than it does de-aggregation/diffusion, ultimately resulting in (4). 4. Initial natural/technical barriers to interworking create private incentives among enough well-placed market participants so that over time the barriers become fixed in place regardless of their original causes, creating an ossified, class-based economy with greatly reduced vertical mobility. 5. Everyone gradually gets used to the new, diminished, less reliable interworking environment, and to the less efficient, less innovative, more risk averse, more stratified, more brittle -- i.e., broadly inferior -- economy that it permits. Bottom line: It's too soon to tell whether TCP/IP will succumb to Gresham's Law, but IMO it's way too soon to conclude that the modest uptick in IPv6 deployment that we're seeing now, finally, means that the risks associated with a Gresham's Law-like adaptation have been eliminated. That said, I heartily encourage my friends and colleagues who are still operators to prove me wrong! TV > However, >> any other outcome would likely arise from and/or push IPv4 >> and incumbent IPv4-based operators into assuming the same >> sort of (economic) roles that have been traditionally >> associated with physical "last mile" facilities and their >> owner-operators... which would ultimately drive every other >> aspiring new routing service provider thereafter into the >> same sort of (basically, adversarial) position with the same >> sort of bypass-centric goals that were/are commonly >> attributed to non-incumbent Internet service providers, esp. >> back in the 1980-1990s. >> >> What other potential options do you envision which might >> provide some reasonable probability of avoiding this >> Catch-22? >> >> TV _______________________________________________ rrg >> mailing list rrg@irtf.org >> http://www.irtf.org/mailman/listinfo/rrg >> _______________________________________________ rrg mailing list rrg@irtf.org http://www.irtf.org/mailman/listinfo/rrg