Did I Say 2010?

I had intended to shift gears this week and write about something other than IPv6, but an update and a couple of very good reader questions compel me to stay on that topic for at least a few more days. First, the update.

I cited Geoff Huston's study on Monday of last week, in which his latest projection for IPv4 depletion of the IANA pool was mid-2011. Geoff updates this study frequently as more data becomes available, and he changed it again the middle of last week-not for new data, but because of observed errors in various predictive models. In previous versions of the study he had based his predictions on an exponential model; but he has now found that a second-order polynomial model is a better fit to the available data. Using that model, Huston's predicted exhaustion of the IANA IPv4 pool changed to December of 2009, and the RIR exhaustion date changed to September of 2010.

Where the Huston and Tony Hain studies previously differed by a little less than a year, they are now very close to coinciding. I suggested in last Wednesday's post that less tangible factors could bring the statistically predicted exhaustion date closer by up to a year; given Huston's new projection and Hain's existing projection, those intangibles mean IPv4 allocations from the IANA are likely to stop in late 2008 or early 2009.

Now on to the reader question.

Mark Seery asks:

The allocation problem is often talked about in terms of different regional impacts, with it often stated, for example, that the crunch not might be as bad in the U.S. as in other regions due to the dynamics of original allocations. Can you comment on whether you think the crunch coming in 2010/2011 will be universal across the globe or staggered region by region?

Very large blocks of IPv4 addresses were allocated to a few corporations and universities involved in original research in the 1970s and 1980s, when few foresaw the current popularity of the Internet and IPv4 addresses were assumed to be abundant. While some have ethically returned unused addresses, other organizations have sat on theirs. And while some Asian countries recognized a coming shortage of IPv4 addresses and began getting aggressive about IPv6 in the late 1990s, North America as a whole has started getting serious about IPv6 only in the last three or so years.

These two factors together have contributed to a widespread assumption that North America, and perhaps to a lesser degree Europe, have more time before their IPv4 supplies will run out.

Here's why that assumption is wrong:

Classless Inter-Domain Routing (CIDR) rules were implemented in the fall of 1993 so that IPv4 address blocks could be allocated more efficiently. Specifically, CIDR did away with the IPv4 concepts of Class A, B, and C addresses and allowed prefix allocations to match actual needs. Part of that new practice was that organizations with existing allocations had to prove that they had used up those allocations before they could get new ones.

Before CIDR, IPv4 allocations were rising exponentially. From 1993 to 1999, the allocation rate was almost flat. And then beginning in 2000 and continuing to today, the allocation rate is again rising exponentially. What that tells you is that CIDR had a significant positive impact on the allocation rate in the 1990s but that by 2000, organizations with pre-1993 allocations had used up those allocations under CIDR rules and were beginning to ask for more.

So while there are still some organizations sitting on unused IPv4 space, these are not the organizations that are creating demand for new addresses. Because they have not given back their unused addresses, they certainly see value in those addresses and if a "black market" in IPv4 addresses starts up they are apt to be players in that market.

Another perspective on this is to look at the RIR allocations over the past few years: In 2004, three of the nine /8 blocks allocated by the IANA were to ARIN; in 2005, four of the thirteen /8s went to ARIN, and in 2006 four out of the ten allocations went to ARIN. These numbers indicate that there is as much demand for IPv4 addresses in North America as in other regions of the world. All in all, the crunch will be felt equally around the world in about the same timeframes.

• LANs / WANs
• IPv4