On Wednesday, I argued that collective action problems will delay the transition to IPv6 for many years, and possibly forever.
The obvious response is that the world doesn’t have a choice. The majority of the world’s population isn’t yet on the Internet and in rich companies the number of devices per person continues to rise. So IPv4’s limit of 4 billion IP addresses has to give at some point, right?
Maybe, but it’s important to remember two points: first, IP addresses can be shared. Indeed, many of them already are. My household has two people and around a dozen devices that share our single IP address. If a decade from now we have 20 or 50 Internet-connected devices, there’s every reason to think that those devices, too, will be able to comfortably share a single IP address.
There is a theoretical limit. Network Address Translation, the technology I described in my last post uses something called a “port number” to disambiguate among hosts on the private network, and the IP protocol allows there to be around 64,000 ports. Very roughly speaking, this means that a single IP address has a theoretical limit of 64,000 simultaneous connections, though a variety of issues make the practical limit lower. In any event, while the number of machines that can share an IP address is not infinite, there’s still a lot of headroom.
The second point is that currently-allocated IP addresses are not being used efficiently. This is an artifact of the era when addresses were plentiful. ISPs could request them essentially for free, and so they didn’t have much incentive to economize. As the official exhaustion point has drawn closer, ISPs have gradually begun to use them more efficiently, but there’s still a lot of room for improvement.
For example, in the Internet’s early days, a number of large organizations, including Apple, MIT, and Ford, were allocated “Class A” blocks of 16 million IP addresses apiece. Apple is a big company that’s probably using hundreds of thousands of IP addresses, but that still leaves plenty of spare capacity it could transfer to someone who needed it more. The reason they haven’t done so, presumably, is that there was no particular incentive to do so. Renumbering a large company’s network is a pain, and as long as ISPs could get IP addresses for free, they had no reason to pay Apple for its trouble
But this is where supply and demand come in. Now that IP addresses are no longer available for the asking, growing ISPs will be increasingly desperate to get their hands on more. Sooner or later, we should expect a market to develop. Apple’s not going to give someone its IP addresses just because they ask nicely, but if someone were willing to pay $10 or $100 per address, then it might be worth Apple’s trouble to go through the hassle of re-numbering its network and relinquishing its addresses.
Network administrators hate thinking of addresses as a scarce resource to be conserved—both because it makes their job harder and because there’s no theoretical reason for addresses to be scarce. But using prices to allocate IPv4 addresses where they’re needed most can extend the useful life of IPv4 for a long time. And given the obstacles to the switch, this may be necessary.
If and when we do eventually move to IPv6, I suspect it will be because the price of IPv4 addresses has risen so high that switching to IPv6 becomes a cost-cutting move. ISPs that make the switch will still need IPv4 addresses to talk to other IPv4-connected hosts, but by routing some of their customers’ traffic over IPv6, they can reduce the number of IPv4 connections per customer and squeeze more customers onto each IPv4 address.
This is likely to happen first someplace like China or India where they have more people and less money than we do in the US. Developing countries joined the Internet late, and as a consequence they already face a more serious shortage of IPv4 addresses.
To be clear, none of this is to say that it’s desirable to stay on IPv4. A shortage of network addresses introduces a number of performance problems and administrative headaches that would be avoided if we all moved to IPv6. But the fact that it’s desirable doesn’t mean it will happen any time soon.
