Peak Internet - Mirage or Malthusian Crisis?
Ken Lefkowitz, Managing Partner
Oct 14, 2016
Is the Internet about to run out of addresses and stop working? This was by far the most fervent discussion at the RONOG3 meeting (www.ronog.ro) that I attended this week. Many engineers out there lose sleep in the belief that it could happen if there's not a massively widespread upgrade very soon of the Internet's 'operating system' from Internet Protocol version 4, or IPv4 for short, to IPv6. (Never mind about IPv5 - kind of like Windows 9, and the Samsung Galaxy Note 7, odd-numbered systems sometimes just don't work out.) There's no Microsoft to run the Internet and hector your ISP into an upgrade with daily reminders, so instead it happens with insistent appeals at industry gatherings such as RONOG.
But is this a real problem? The debate reminded me a lot of the peak oil controversy (remember peak oil?). In 2007-08 when the price of oil was soaring, oil companies were exploring all kinds of exotic and expensive resources such as tar sands and deep offshore. People were worried that because oil resources are finite, all the hydrocarbons in the world worth discovering had been found, and oil production would not keep pace with demand. This would trigger an out-of-control price spiral and leave the world economy to literally run out of gas. But today we are awash in an oil glut, with prices at around 1/4 of the 2008 peak. Well guess what: there are currently more than a billion unused IPv4 addresses and they trade well below €10.
The classic statement of this type of problem is from Thomas Malthus in his 1798 piece, An Essay on the Principle of Population. Malthus argued that human population grows exponentially while agricultural output grows arithmetically. He did the math and extrapolated a population/famine crisis. What actually happened was the Industrial Revolution -- people started using resources much more efficiently -- making a mockery of his forecast.
The point is that people, especially hungry people, change their behavior to get their needs met - they transmit and respond to price signals and they re-allocate resources accordingly. "It's economics, stupid," to paraphrase Bill Clinton. If the price of food jumps up, people plant more crops. If the price of oil jumps up, oil companies start combing the ends of the earth for more of it. People also build and buy more efficient cars.
It follows that if IP addresses are scarce, people will stop wasting them and will make more of them. Now perhaps the main reason for the development of IPv6 was to make more addresses - problem solved, right? But here is the crux of the debate: changing systems or technology is costly.
If you're old enough to remember the oil crises of the 1970s, people were evangelizing electric cars way back then, but only now, 40 years and another crisis later, are EVs creeping up the adoption curve and becoming competitive in terms of price and features with gas and diesel cars. It was much easier to improve fuel-economy standards of traditional cars. Likewise with IPv4: Ciprian Nica from IP Broker made the point that if you're a large cable TV operator, why spend €100 on an IPv6-capable set-top box (and multiply that cost over millions of subscribers) when you can buy an IPv4 address for a cost that is an order of magnitude less from somebody who was just sitting on that address? Without the large carriers on board, there will be no critical mass for IPv6 adoption.
That triggered an outcry: "but you have a commercial interest!" Yes, for an IP address broker to say this is blatantly self-serving, but the point is not any less valid just because an interested party makes it. Commercial interest is a blind spot for too many engineers: IPv6 is a great solution whose time hasn't yet come because there are dollars and cents obstacles to adoption, and users (and operators) respond to those price signals by changing their behavior.
This blind spot is evident in other areas as well - take BGP, the protocol that runs the guts of the Internet without taking into account the cost of a given route or its physical distance. When I try to point that out to network engineers, I typically get a funny look implying 'who is this wise guy who doesn't know how to configure a router trying to tell me about routing?'
After close to a couple of hours of back and forth about the facts, Ciprian, who was in the minority, defended the point well enough to push the discussion forward. Jan Žorž from Internet Society noted that IPv6 has features that allow great new things to happen on the Internet. To me that sounds like the pivot point that IPv6 promoters can actually use to lever up adoption from its current 15% of Internet traffic. If free addresses are not a driver for adoption, then what about those features? What is the killer app -- the Tesla Model S of IP -- that will end IPv6's nearly 20 years in the desert and bring us home to the promised land? Maybe we'll come to that at the next meeting.