Slide from a presentation on the exhaustion of the supply of addresses for devices on the internet (pdf) by Geoff Huston, Adjunct Research Fellow at the Centre for Advanced Internet Architectures
Seven days from today, according to ISP Hurricane Electric, the organization that gives out the unique 9-digit addresses that in theory identify every device connected to the Internet is going to simply run out of those addresses. (This is only an estimate; Japanese ISP iNetCore pegs the end to 2 days later.)
It’s as if the local telephone company simply ran out of phone numbers to give to its customers. Except every time this happened, back when everyone had a land line and area codes meant something, the telephone company saw it a mile away, and added a new area code for a given geographic area, thus averting catastrophe.
In this case, there is no plan B. Years – decades, really – of foot-dragging mean that the world’s hardware manufacturers, OS coders, website builders and Internet service providers are stuck with the existing system. When the last block of IP addresses, as they’re known, is handed out, that’s it.
Does this mean no more computers can be connected to the Internet after January 31st, 2011? No, of course not - if you’re Apple, HP, Xerox, Ford or one of the other companies that got a gigantic block of IP addresses in the early days of the Internet, you’ll hardly notice; these companies are using only a fraction of the IP addresses they reserved all those years ago.
Everyone else, the rabble drawing from the well of free IP addresses that’s about to run dry, has problems that will slowly grow worse. At first, your Internet Service Provider is going to solve the problem for you, possibly by buying more IP addresses for what are likely to be ever higher prices as the gap between demand and supply widens. What’s mostly likely, though, is that your ISP is going to implement a kludge: they’ll use something called Network Address Translation (NAT) to hide more and more users under one of the IP addresses they already possess.
In other words, you’re going to start sharing a phone number with a stranger. If they misbehave and are banned, you’ll be banned as well. That’s unlikely, but here’s what’s almost assured: slowly, our experience of the Internet will start to degrade. If we’re lucky, the change will remain imperceptible, and if we’re not, well…
As Google engineer Lorenzo Colitti recently told Agency France Presse:
“You will start to share with your neighbors, and that causes problems because applications can’t distinguish you apart,” Colitti said. “If your neighbor ends up in a blacklist, you will too.”
“The Internet won’t stop working; it will just slowly degrade,” he continued, explaining that systems would eventually have trouble handling multiple connections on shared addresses. “Things will get slower and flakier.”
Though NAT has been in use for a long time, most applications are tuned to pass their data through only a single layer of NAT. And that single layer is almost certainly already in use – NAT is the reason that every device in your house connected to your WiFi base station appears, to the Internet at large, to have the same IP address.
What happens when your BitTorrent client or your Skype or other VoIP application starts communicating through more layers of NAT, which works by shunting data through other ports, some of them less desireable, is that things break down. Generally, the Internet is OK with that – packets of data can be re-sent when they fail to arrive – but at the cost of speed.
It’s not as if the Internet is going to break in a week. The correct analogy is rot - in a week, the shipworms are unleashed, and the planks that stand between you and the briny depths of an unusable connection to the most important communications network on the planet begin to be compromised by a potentially endless series of imperfect hacks designed to postpone the inevitable.
The inevitable, of course, is an upgrade from the current Internet communications protocol, IPv4, to the next-generation standard, IPv6. The problem with that inevitability is that the price of switching to IPv6 is going to be so high - in terms of reliability, backwards-compatibility, actual money and a hundred other potential issues, that before the bulk of the users of the Internet are finally forced to switch to IPv6, the existing IPv4 network will probably have to degrade to some extent.
Speculation on how bad things will get before that happens is pointless – it’s a question of switching every device in the world over to a new protocol; in essence, an economics question. And we all know the value of predictions in that realm.