Today a new /24 (formerly termed a Class C) block of 256 addresses is allocated to somebody every 30 seconds. At that rate, the Internet loses an entire /16 block (a Class B - 65,536 addresses) every two hours. That's 24/7 friends, and the rate is accelerating. Based on the current rate of growth, which Hain documents extensively in his report, we'll have allocated all IPv4 address space in just half a decade.

This prediction adds a new sense of urgency to the IPv6 migration debate. IPv6, which uses 128-bit rather than 32-bit addresses, provides essentially unlimited IP space. But switching the entire Internet over to IPv6 is an expensive undertaking, and many IPv6 opponents have argued that the original 2022 prediction meant that IPv6 is still far off. Some (whose initials are NSF) even say IPv6 is no longer necessary, proposing that we go back to the drawing board and design yet another IP addressing scheme, given that the IPv6 design is already 10 years old.

Hain says we don't have time for that, and his argument is persuasive. He uses the actual IP address allocation history from the Internet Assigned Numbers Authority (IANA) in his analysis, points out the reasons for the APNIC's previous rosy outlook, and explains why those reasons are no longer valid.

I'm often leery of statistical predictions because there are so many ways to lie with statistics. However, Hain bends over backward to be fair and transparent in his reasoning, and even gives the author of APNIC's report, Geoff Huston, a chance to respond in the same document. Huston's answer boils down to the two predictions being within the margins of statistical error. However, the difference between the two dates is significant: moving to IPv6 within five years means everyone needs to begin planning their migration strategy now.

Both authors agree that market forces will automatically curtail IPv4 growth at some point, by making IP addresses more and more expensive as the end approaches. Both also agree that this effect will not be good for business, and in the long run will be more expensive than the cost of moving to IPv6.

Both writers also point out that predictions presume that growth conditions remain as they are, which is almost certainly not likely to be true for the remainder of IPv4's lifetime. Already new network applications, such as mobile computing, are hungrily chewing through IP addresses. And many third-world countries have yet to get much of their populace on the Internet. Both of those foreseeable demands could exhaust IPv4 space even sooner than five years.

Conversely, there could be a growth-slowing technology collapse like the dot-com crash, but given the business lessons learned since 2000 that seems a foolish bet.

The report includes a lively and fascinating roundtable discussion between Hain, Huston, and Internet gurus John Klensin and Fred Baker. All participants concur that IPv4 has entered its end game, and the time has come to get serious about IPv6 migration.

Tony Hain's report A Pragmatic Report on IPv4 Address Space Consumption:
http://www.cisco.com/en/US/about/ac123/ac147/archived_issues/ipj_8-3/ipv4.html

Geoff Huston's APNIC IPv4 Address Space Report:
http://bgp.potaroo.net/ipv4/

Here's a bonus goody. APNIC produced a movie showing the actual consumption of IP addresses over time, based on Internet routing table data. It's a fun, and frightening, look at the rate of IPv6 network growth:
http://www.potaroo.net/avi/comp.m1v