[arin-ppml] Encouraging IPv6 Transition (was: Clarify /29 assignment identification requirement)
tvest at eyeconomics.com
Sun May 13 02:04:24 EDT 2012
On May 13, 2012, at 12:59 AM, Owen DeLong wrote:
>> Hi Owen,
>> Though I am in wholehearted agreement with both of your general points (about markets and open source), as well as with your your specific (counter)examples, I think that it is worth contemplating whether TCP/IP might never had achieved critical mass in the absence of the unique distribution mechanisms and principles that were ultimately, roughly codified in RFC2050 and institutionalized in the form of the "RIR system." My own interpretation of the historical record suggests that the interests who favored different, proprietary routing and addressing technologies (including both the proprietors themselves as well as the established telecom SSOs where their views went unchallenged), and who at that time embraced TCP/IP as, at best, an optional inter-networking feature that could be superficially bolted onto their own proprietary networking platforms, were still dominant almost everywhere except the United States at the time when the first RIRs appeared in Europe and Asia. I believe that the limited geographic devolution of the critical protocol resource distribution and policy-making authority that began with the establishment of RIPE NCC substantially reduced both the number and severity/cost of barriers to establishing and expanding a TCP/IP-based network -- rather than, say, a DECnet, SNA, or some other proprietary network -- and that those relative advantages played no small role in determining TCP/IP's subsequent path to absolute global market dominance.
> I absolutely agree. However, I'm not sure how that differs from what I was saying, if you believe that it does.
> To me, it seems like another way of saying "not only is a market likely to be harmful to what we have achieved here, but, had we started with a market, we would likely have never achieved widespread adoption.
>> Conceivably, one could argue that TCP/IP would eventually have achieved that same dominant status by some other path -- e.g., if the inherent superiority of TCP/IP protocols, or the growing market power of US-based, TCP/IP-favoring hardware manufacturers and/or network services providers would have eventually proven irresistible even in a world in which TCP/IP was just a parochial US-only standard -- despite the fact that prospective TCP/IP adoptees would have understood that choosing TCP/IP would make them perpetually dependent on some US-based entity -- either a USG contractee, or more likely GE, Halliburton, or one of the other early private entities who obtained classful IPv4 before its c. 1994-1995 exhaustion. Or perhaps HP, Dell, IBM, et al. and the telco-era SSOs would have eventually embraced "open access" approaches toward inter/networking protocol resources than were similar to or even stronger than the "needs based" practices pioneered by RIR system participants -- or barring that, maybe a whole new generation of open source-favoring hardware manufacturers would have inevitably sprung up somewhere, and inevitably competed all of the dominant, closed-network/protocol-favoring manufacturers and service providers into submission, or bankruptcy...
> I think that by and large, TCP/IP won because it came with zero vendor baggage. Any vendor was free to implement it without fear of giving a competitive advantage to their competing vendors. No other protocol offered this possibility. While the moneyed interests were definitely carrying on their proprietary protocols,
> to be sure, they were faced with groups of companies that wanted interoperability between their networks no matter whether they bought from Novell, Banyan, DEC, or whoever. As a result, each of them was willing to bolt-on TCP/IP as this optional add-on (sometimes at significant cost) to enable those minimal inter-company connections where necessary.
> Where I think that the InterNIC and later the RIR system triumphed is that it presented a low-friction low-cost mechanism for distribution of addresses (RIRs actually added, not reduced friction and cost vs. InterNIC, btw) that still provided reliable global uniqueness. Had the InterNIC and later the RIR system not succeeded in that mission, TCP/IP would likely be several different fragmented administrative zones with odd NAT boundaries connecting them.
Sounds like we're in complete agreement on the main points. The specific point that I was trying to highlight (the independent significance of which we may differ slightly) was that, contra the markets-fix-everything comments that preceded yours, TCP/IP would probably have looked either a lot less useful or a lot more vendor baggage-laden (or both) by the mid-1990s -- that is, if "needs based" distribution mechanisms had not been formalized and geographically distributed around that time. I have no doubts about the greater ease of interacting with InterNIC relative to its successors (esp. for English-speaking residents of adjacent timezones), and I know that opinions still differ among some who were more than just present at the creation regarding the merits of the geographic part of that equation. That said, from what I know of the relevant chain(s) of events outside the US, I would venture that whatever marginal friction has resulted from the regionalization of number resource management pale into insignificance when compared to the frictions that would have been endemic in a world without open access to open source interdomain routing protocols. IMO, given what was at stake and the uncertainty of the outcome at that time, it still looks like it was the smartest bet.
>> Anything's possible I guess, but speaking as someone who was born in the 20th century, and who has benefited a great deal from the existence of a working Internet during my lifetime, I'm very glad that we didn't have to wait to find out whether one of those other highly improbable sequences would have eventually panned out.
> I couldn't agree more, but, having lived as a network engineer through much of that history and transition, it seems we have divergent memories of the exact chain of events.
That's okay Owen; we have plenty of time to work on that ;-)
More information about the ARIN-PPML