When the Internet Runs Out of IP Addresses

Experts predict that within three years we will see the last of new Web addresses. What will happen then? The best solution is to create a market for already assigned but unwanted numbers, says Harvard Business School professor Ben Edelman. Key concepts include:
  • Internet regulators are running out of IP addresses to assign new Web-based sites and services.
  • The next IP standard is probably years away from widespread deployment.
  • In the meantime, creating a market for unused addresses could help bridge the gap.
by Sean Silverthorne

The Internet is running out of room.

Experts predict that in two or three years we will run out of Web addresses, so-called IP addresses, that can be assigned to new Internet-based sites and services.

Each site is assigned a unique number based on the IPv4 standard. IPv4 is the basis of addresses using combinations of four integers, about 4 billion possible combinations.

The problem: Internet growth is so dramatic that potential IPv4 addresses are running short on supply. A new standard called IPv6 will solve the problem with trillions of possible combinations, but technical and practical roadblocks are delaying its widespread implementation, probably for many years.

So what happens when the last IPv4 address is assigned? Harvard Business School professor Benjamin G. Edelman proposes a solution: Create a market for holders of previously assigned but unused addresses to sell or otherwise transfer them to new owners.

"It's unlikely that other networks would return their space for free—why would they?" says Edelman. "But if the price is right, they may be willing to transfer the space to someone who needs it more."

Our Q&A follows.

Sean Silverthorne: Why are we running out of IPv4 addresses?

Ben Edelman: In this respect, the Internet is a victim of its own success. The IP address system was never designed for the large, complicated, widely used Internet we enjoy today. We're fortunate in that we've gotten this far without having to upgrade the Internet's addressing fundamentals. But the time is coming when such upgrades will be unavoidable.

Usually when a resource becomes scarce, its price increases. But that hasn't happened here because IP addresses are provided at de minimis cost—a regulatory decision reflecting that IP addresses are ultimately just numbers, and that it's odd to have to pay for a number. With prices stuck very close to zero, and demand steady and growing, economic incentives invite exhaustion.

Q: What happens if nothing is done and we run out of addresses? Does the Internet stop growing?

A: If networks can't get new IP addresses, it will be much harder for them to grow. They have some options, like making more intensive use of the addresses they already have through address sharing (formally, network address translation, or NAT) or perhaps revisiting which addresses truly need IP addresses. (Does that laser printer actually need a globally unique address? Maybe it could make due with a local-only address.)

But these approaches have major challenges. For example, NAT impedes many kinds of applications, like videoconferencing and certain file transfers. Large-scale NAT will limit innovation, increase complexity, and ultimately make the Internet less useful than it could be.

Another important challenge is that if networks can't get new IP addresses, it will be harder to enter many technology businesses. Want to start a new ISP? Or run a service provider that hosts a large number of Web sites? Without ample IP addresses, it's difficult to get started in these businesses. Entry and potential entry are an important part of competition. We need to make sure new firms can easily begin operations so that existing providers can't hold customers hostage.

Q: Isn't a transition to Ipv6 supposed to solve this problem?

A: IPv6 offers important benefits. In particular, if all networks ran IPv6, there would be plenty of addresses for everyone, and we'd have no further address shortage.

But it's difficult to get from here to there. Right now, users all run IPv4 computers, Web sites all run IPv4 Web servers, and ISPs all offer IPv4 transit. Transition raises the question of who goes first. ISPs have hesitated to offer IPv6 service because there's not much demand. Users (and end-user networks like companies and universities) aren't demanding IPv6 because, at least for now, they still can get all the IPv4 addresses they need. Neither are Web sites demanding IPv6, because Web sites want to be reached by users, and users only run IPv4.

In fact, it's even worse than that. Moving to IPv6 too early brings extra costs. For one, there are the standard costs of learning how to set up IPv6, and allocating engineering time to handle all the details. But when a Web site supports both IPv6 and IPv4, some users will mistakenly try to reach the site by IPv6 because their computers and network cards are misconfigured. (For example, certain security enables IPv6 in order to install IPv6 security—which seems like a good idea but actually results in v6 being enabled when it shouldn't be.)

These problems don't affect that many users—measurements suggest a fraction of a percent. But that's enough. If you're Dell, would you want to turn on IPv6, with no current benefits and an immediate loss of a fraction of a percent of your users? It just doesn't make sense, especially when your competitors can stay with v4 alone and be reachable by 100 percent of users as always.

Q: You advocate a market-based solution, at least as a temporary measure. Can you describe the idea of transferring unused IP addresses?

A: The basic idea is that some networks have ended up with more IPv4 addresses than they need, while others have less than they need (or will need in the near future).

Why do some networks have more than they need? A variety of reasons: Some networks received extra-large allocations of 16-plus million addresses in the Internet's early days, when addresses looked abundant and when it was hard to subdivide address blocks into just the right sizes. Other networks have scaled back their IPv4 needs through a change of business focus, or even through bankruptcy.

Q: What are the benefits of a market-based plan? How much time would this buy us?

A: A market-based approach offers a real benefit to those who still need more IPv4 addresses after ordinary supplies run out. Rather than being told that no more IPv4 space is available, on any terms or at any price, these networks could offer payments to get v4 space from others. It's unlikely that other networks would return their space for free—why would they? But if the price is right, they may be willing to transfer the space to someone who needs it more.

So the core benefit is allocative efficiency, moving scarce resources to those who need them most.

But there are other benefits, too. By putting a positive price on IPv4 space, a market mechanism would remind current v4 users that their v4 space is valuable, and that they might want to try to vacate it, to the extent they can, perhaps by moving to IPv6. A market basically tells networks: "We will pay you to use v6 instead." That's a transition incentive quite different from anything we've seen to date. That's a transition incentive that just might work.

Q: Wouldn't speculators take advantage to drive up prices?

A: Some people are definitely worried about speculators. But I don't think they're going to be a big problem here.

For one thing, the proposed transfer rules are slated to require that an address recipient be a bona fide network that actually needs, and can use, IP addresses. The responsible regulators, like North America's American Registry for Internet Numbers (ARIN), have been making these determinations for more than a decade. They're going to keep performing these reviews even when the resources at issue are resources obtained via paid transfers, rather than resources obtained from as-yet-unused space.

Furthermore, the dynamics of this market will probably make it unappealing to speculators. In the long run, the Internet will move to IPv6. So anyone speculating on IPv4 knows the price is going to plummet, probably to zero, in due course.

How soon will that be? That's much harder to say. This is a one-of-a-kind transition, importantly different from the various other standards transitions we've faced before. For example, the digital TV transition benefited from a strong central authority, the U.S. government, that could order transition on a particular date certain. Not so here, for there's no one to order ISPs to run v6 rather than v4. So speculators would try to predict the future at their peril.

Q: If your plan was adopted, would consumers pay more for Internet access and to develop Web sites?

A: If this transition goes smoothly, consumers should never notice. To date, IP addresses have been a trivially small part of the cost of Internet access and Web site hosting. Even if IP address prices increased 100 times, consumers still probably wouldn't notice.

The bigger worries come if ISPs just cannot expand, or just cannot enter the market. If that were to come to pass, I wouldn't be surprised to see effects on service price and quality. That's why it's so important to make the transition smooth—to provide financial incentives to move to IPv6, and to provide a framework to let ISPs enter and expand even their IPv4 operations at reasonable cost and with appropriate predictability.

Q: Who will make this decision? Will governments play a role?

A: IP addresses are given out by five Regional Internet Registries (RIRs). In North America, our RIR is the American Registry for Internet Numbers (ARIN). RIRs are private nonprofits, not a government agency, and their powers are appropriately limited. But RIRs are in a position to allow paid transfers, if they conclude that such transfers are in the Internet's best interests.

Governments definitely play a role. For example, the U.S. Office of Management and Budget (OMB) mandated the installation of IPv6-capable equipment on agencies' network backbones by June 2008—certainly a step in the right direction, and one of the few major drivers of IPv6 deployment. But so far OMB hasn't yet tried to force agencies to actually use IPv6, knowing that it would be too costly, too inconvenient, and on balance too hard to justify, at least for now.

More generally, the Internet's infrastructure is largely private. ISPs make their own decisions about what systems to install, and what services to provide. Certainly governments can offer incentives, but when Japan experimented with IPv6 deployment incentives at the start of the decade, the payments had limited effectiveness.

So at present, I think the Internet's best hopes come not from governments or outside deployment incentives, but from internal incentives grounded in putting a positive price on increasingly valuable IPv4 resources.