But guess what: The Internet has grown so large that it’s now pushing up against some of the limits of its original design. The moment is fast approaching when it will no longer be able to incorporate additional hardware: No more PCs, no more servers, no more routers, no more mobile phones. Unless something changes deep inside, the Internet will simply cease to grow any more, and only those machines and devices lucky enough to be attached already will enjoy its many benefits.
Or maybe not. All this is the subject of a big debate in certain information technology circles these days. In fact, a major update to the Internet’s foundations, an update designed to enable the Internet to grow billions of times over, is already in place, just waiting for the switch to be thrown.
The gist of the problem is this: Each piece of hardware attached to the Internet must have a unique numerical address, assigned to it by one of the authorities that governs the Internet. This set of numbers—generally hidden from view behind a URL, is similar to area codes and telephone numbers that enable calls to take place between telephones. These addresses are key to the movement of mail, Web pages, videos, and other data between the Internet’s millions of machines. But due to the format of the Internet’s addressing scheme—namely, the fact that the number of digits in each address is fixed—the number of unique addresses available is finite.
The Internet’s addressing scheme is defined as part of the Internet Protocol (IP), which describes exactly how to assemble the information in each packet of data sent through the Internet. One of the fields of data therein is the address of the packet’s intended destination, and in the current version of IP, called IPv4, this address field contains 32 bits. That’s enough to create approximately 4.3 billion unique addresses—a big number but not nearly enough, it appears, to handle the flood of new devices that people want to attach to the Internet.
The world’s several hundred million desktop and laptop PCs and Web and email servers turn out to represent only the start of the Internet’s growth ramp. Also potentially in need of their own addresses are potentially billions of mobile phones. And in the wings are tens to hundreds of billions more electronic devices, from home DVD players to refrigerators and individual light switches to every kind of industrial machine—all potential objects of remote monitoring and control—and all needing their own IP address.
Fortunately, the possibility of so many billions of new devices joining the Internet was envisioned many years ago, when a much-advanced version of IP, known as IPv6, was worked out. IPv6 includes an address field of 128 bits, which in theory makes available 2 to the 128th power, or 3.4×10 to the 38th power unique addresses. That’s enough for each of the 6.5 billion persons living today to have 5x10 to the 28th power IP addresses of their own.
So why, then, a decade after IPv6 was defined and with IPv4 addresses running out, is there such little interest in the new protocol? In fact, millions of PCs and Macs have IPv6 built into their operating systems, but usage remains extremely rare among the Internet’s larger host computers, which run Web sites and so forth. According to a survey conducted by Google, only about 1% of hosts are using IPv6, but without significant host participation, the new protocol doesn’t help anyone.
One reason for the lag is that, as often happens with technology, the networking industry has worked out alternatives that have enabled IPv4 to handle more devices without a wholesale changeover to IPv6. Techniques called Network Address Translation (NAT) and Dynamic Host Configuration Protocol (DHCP), for instance, enable Internet Service Providers to share a relatively small set of IP addresses across many customer computers. These techniques don’t scale forever, though, and have difficulty handling peer-to-peer (P2P) connections, so the conversion to IPv6 looks inevitable. And the sooner, the better.
How to speed the transition? One approach might be to create a market in IP addresses, which today are essentially free for the taking. Once the price for increasingly scarce IPv4 addresses starts rising, the theory goes, buyers would naturally opt for IPv6 as a way to save money. In analyzing this idea, Benjamin Edelman, an assistant professor at Harvard Business School, points out that measures likely would be needed to discourage rampant speculation in addresses and to help with certain technical issues that arise inside the Internet’s routing structure as IPv4- and IPv6-based traffic coexisted for a period of time.
But in a widely-circulated working paper published in early 2009, Edelman says, “Experience in other markets indicates that trading resource rights can achieve large efficiency gains. For example, tradable pollution rights reportedly reduced pollution for 55% lower cost than ordering across-the-board cuts by all firms thanks to variation in firms’ costs of abatement. Networks feature similar variation in their initial address allocations, in the suitability of transition technologies to serve their requirements, and in their staff and equipment costs of migration. Through transfers, networks with low transition costs can move to v6 first—at lower cost than transitions in random order.”
Not everyone is convinced. As it turns out, market mechanisms for allocating scarce IPv4 addresses have been proposed for many years. They’ve not caught on, experts say, because markets are viewed as helping only for a few years; because ownership of IP addresses is explicitly prohibited by key Internet policy documents; and because trading in addresses would tend to fragment the Internet’s so-called “address space” and create severe technical and service complexities. In short, ISPs would be better off paying for the conversion to IPv6, some observers say, than to constantly reshuffle IPv4 addresses as they buy and resell those.
Exactly when the IPv4 address space will be fully exhausted is difficult to predict. A few years ago, 2009 was given as the date, but increasing reliance on NAT gateways and other IPv4-based workarounds has helped ISPs to push the date back. What’s more, large blocks of unused addresses still remain in the hands of government agencies and mounting political pressure could free those addresses up for use by the private sector. But everyone agrees that demand will continue to rise and it’s clear to all involved that by 2016, IPv4 addresses will certainly be all used up.
Fortunately, the old and new versions of IP can co-exist, as they have been for several years now. But eventually, system-wide adoption will be necessary. Several Asian countries already are well ahead of the United States in bringing ultra-high-speed Internet service to homes—100 Mbps service is common in Singapore and Hong Kong, for instance—and they are leading the way in IPv6, too. A major showcase for IPv6 was the 2008 Summer Olympic Games, held in Beijing. If only to keep its national economy competitive, the U.S. will have to find one way or another to speed adoption of IPv6.
To read more case studies, visit Think+Up.




