The Internet Just Ran Out of Numbers
Every available Internet protocol address has now been spoken for—which could mean huge trouble down the line.
On February 3, it finally happened: the clock ran out on the Internet as we know it. That was the day that the stash of Internet protocol addresses that are used to identify and locate computers connected to the Internet—the telephone numbers of the online world—was exhausted.
The problem is that the current system for IP addresses, IPv4, uses numeric addresses that are 32 bits long—giving a total of just over four billion potential numbers, which must have seemed like a lot when IPv4 was introduced in 1981. But there are now seven billion people on Earth, and more and more of them—and their devices—are going online all the time. Fortunately, engineers realized the limitations of IPv4 a long time ago and lined up a successor, called IPv6, in 1998. (IPv5 was an experimental system that never went public.)
IPv6 uses 128 bits rather than 32, producing a pool of numbers that is staggeringly huge—some 3.4 x 10 to the 38, or 48 octillion addresses for every person on Earth. The trouble is that although most servers and all major operating systems have adopted support for IPv6, Internet service providers have been agonizingly slow to follow suit.
For ISPs, it’s a straightforward business dilemma: the two addressing schemes are not directly compatible, which means it would take a significant investment to let IPv4 users connect to IPv6 services. And having relied on the same system for as long as 30 years, they may not feel the need to change.
“It really highlights the failure of the Net at the most basic level to innovate, despite the fact that at the visible levels, it has had unbelievable innovation,” says Jon Crowcroft, Marconi professor at the University of Cambridge’s computer lab.
He points out that the current concerns about IPv4 space don’t really affect those who already have an address—only those who need new numbers. So it is a minor problem for ISPs that have already stockpiled blocks of IPv4 addresses.
“Why does anyone with IPv4 space care? It’s all working, and there’s been no big, terrible disaster,” Crowcroft says. “But it will be interesting to watch how this slow degradation of things [affects] new entrants.”
“New entrants,” in this case, could mean nations with rapidly expanding online populations. Such countries may face significant trouble if their allocation of IPv4 addresses fails to keep up with their appetite for connectivity. Countries like China are already beginning to concentrate on IPv6 support, with the result that parts of the Internet are being created that are, effectively, inaccessible from the parts of the world that only use IPv4.
While the idea of Internet balkanization might sound disturbing, in practice this is still not a pressing issue for ISPs in the West. There is, however, one area where Western nations might begin to feel the squeeze: the “Internet of things.”
The Internet of things is a vision of a world where many more devices can, and will, be connected to the network. Many of us are already familiar with ecosystems of interconnected devices—computers, printers, mobile phones, and even TV sets—that each have their own identity and yet all exist as individual nodes of a wider system.
The Internet of things takes that concept several steps further: it suggests that almost any object—potentially every manufactured object on the planet—could one day have its place in this system. Advocates foresee a world where everything from your clothes to your car to your cup of coffee can be uniquely labeled as a node on the Internet.
Why? Because with the Internet of things, if you lose your keys, the network tells you where they are. Your running shoes tell you when they’ve gone past their optimum mileage the second it happens. Businesses would be able to tell where every product they sell is located. Farms could use irrigation equipment that “talks” to soil sensors to determine how much water is required in each part of a field.
It might sound extravagant, but the shift toward such a world has already begun.
Because of rapid mobile adoption and the spread of technologies such as radio-frequency identification, Ericsson Labs predicts that 50 billion connections will be required by 2020—tough to achieve under IPv4 but well within the reach of IPv6.
But even with the looming Internet of things, IPv4 may still stick around. Even though all IPv4 addresses have been allocated, they aren’t all active. We could see secondary markets for address space develop, particularly among those businesses and universities which—typically by accident—own vast chunks of IPv4 space that go largely unused.
There are other ways to keep IPv4 viable for some time. A technical solution such as network address translation, for example, takes a single public IP address and splits it among many private addresses— allowing devices inside, say, a home or office network to connect to the Internet without their own unique IP addresses.
So even if IPv6 remains out of favor with ISPs, the Internet of things may still arrive. That will please its fans, but should not calm their fears entirely. After all, says Crowcroft, choosing inelegant solutions today will come with costs further down the line. “There are lots of workarounds, and we can do more of that,” he says. “The big problem is that when things go wrong, debugging the Internet is a bitch.”
Become an MIT Technology Review Insider for in-depth analysis and unparalleled perspective.Subscribe today