As we run out of IPv4 address space, is it time to create an exchange for trading unused address blocks? Ars contributors Iljitsch van Beijnum and Timothy Lee tackle the issue. In this article, Iljitsch explains why this is a bad idea. You can read Tim's take here.
The Internet is a packet-switched network. That means all communication going across the network is put in packets, which are transmitted individually. This has the advantage that there's no call setup overhead, like in connection-based networks (think landline phones). But the downside is that each of those packets, holding not much more than one kilobyte of data, must be routed through the network individually. So a big router that handles many millions of packets per second has to take the destination Internet Protocol (IP) address from each packet and then walk through its routing table to find where next to send the packet. This makes the design of routing table data structures and the algorithms to search through them an extremely critical part of the Internet.
The design is so critical that it's necessary to limit the way in which addresses are given out so routing tables remain small and efficient. Hence the earlier limitation that IPv4 addresses could only be given out in class A, B, and C blocks, and the current limitation that block sizes must be powers of two. This was a traumatic change made in 1992-1993, allowing the Internet to survive a crisis that could have killed it as routing tables were quickly outgrowing the capacity of the day's routers.
At around the same time, the first three of the eventual five Regional Internet Registries were formed, which took up the task of distributing IP addresses in North America, Europe, and the Asia-Pacific, respectively. At this point, the policy "to each according to his needs" was made explicit, and organizations requesting address space had to sign a contract spelling out that "IP addresses aren't property."
Interestingly, during the following decade when the address allocation policies were a bit more liberal on paper (and a lot more in practice), and the address registration fees were low, few people, if any, argued that addresses should be a tradable commodity. That all changed when the end of the (almost) free IPv4 address supply drew nigh and IPv4 addresses started to become scarce enough to hold monetary value. Of course, most of the organizations that received IP addresses before the RIR system was set up never returned them, or even signed a maintenance contract with the RIRs. And the RIRs never pushed hard—read: sued—to reclaim legacy addresses. Worse, 268 million IPv4 addresses will probably remain unused forever because pretty much all operating systems and network equipment vendors have a line of code somewhere that says "addresses above 240.0.0.0 are reserved for future use—you can't use those!"
In the meantime, everyone dragged their heels installing IPv6, which handily solves the current IPv4's address scarcity and then some. The current situation is that Asia is already out of IPv4 addresses, save for one last block of 1024 per Internet service provider. Europe will be in the same situation at some point in 2012, or maybe even later this year, with North America to follow one or two years later. IPv6 deployment is growing, but nowhere near fast enough to pick up the slack, partially because of the same lack of foresight mentioned earlier and partially because of a vicious circle of lack of demand/supply/real-world experience.
So with the RIR supplies of IPv4 addresses running low and their policies getting more draconian as the end nears, is IPv4 address trading the magic bullet that will buy us time until IPv6 deployment is of sufficient scale?
Yes and no. But mostly no.
Unpredictability and fragmentation
Let me start with a fairness issue. The US holds about four IPv4 addresses per capita, and most of Western Europe has one or two. But many other parts of the world have much less than that. China has been using up IPv4 address space like it was going out of style, going from having about 8 million addresses around the turn of the millennium to 330 million now. India, on the other hand, has almost as many people but only 35 million IPv4 addresses. Should the poorest countries in the world be forced to buy IP addresses from the West, providing a windfall to some of the richest American companies just because those participated in an e-landgrab at the right moment?
Fairness aside, it's important to recognize what we're talking about. If you have a website to host, you really only need a single IP address. Or maybe even a fraction of an address—you can often host multiple websites on a single IP address. So paying $10, $100, or even $1,000 for an IPv4 address is not a big deal. When I checked a few years ago, the dozen or so biggest service providers in the ARIN region (North America) used up 85 percent of the addresses given out that year. Hosting services doesn't use much address space in the grand scheme of things, and can be made to work in a variety of ways, with or without address trading.
But what about those big service providers? When Comcast needs another 8 million addresses, will it buy those for $10 a piece? Suppose that it does. Then one or two years later, how much will the cable giant have to spend as supply of addresses gets tighter? Will it spend hundreds of millions of dollars on IPv4 addresses every year, with the prospect that this cost will just increase year over year? And worse, as the largest blocks find a new home, large service providers will have to buy up more and more small blocks to secure the same number of addresses. Just dealing with so many different contracts will be a nightmare.
So my prediction is that at the ISP level, a functioning market won't form at all, or will break down very quickly after it forms. All the while, address trading will take away resources, monetary and otherwise, from implementing the long-term solution: IPv6. With no new supply of IPv4 addresses and an increasing number of potential address users (we've been using up 200 million new IPv4 addresses per year recently), an address market will be prone to bubbles. Bubbles can also easily burst as service providers move to address-conserving technologies such as NAT and IPv6. Of course the US government can always bail out service providers using the nearly 200 million legacy IPv4 addresses that it has on its books.
If address trading happens in non-trivial volumes, the address space will also fragment as organizations sell off only part of their address space. Due to the power-of-two limitation, doing this easily adds a handful of entries to the routing tables of routers throughout the world. The IPv4 Internet will become less reliable as older routers operate at peak capacity and routing protocols are stressed. Unless the RIRs throw all their policies out the window and rubber-stamp all address transfers, it will become harder to trace back an address to its user, giving free reign to spammers and other shady outfits.
Internet users are about to encounter a big pothole in the information super highway. What they need to clear this hurdle without breaking an axle or worse, is predictability. Neither the RIRs tightening the supply of the remaining IPv4 addresses nor shady markets where exchanges happen hidden from the public eye that keep IPv4 running for indeterminate amounts of time are helpful here. But the good news is that IPv6 addresses can be had without conforming to oppressive policies and essentially for free. Ask for some at your friendly neighborhood service provider.
Listing image by Photo illustration by Aurich Lawson
reader comments
100