The Scarcest Internet Resource?
Saturday, January 9th, 2010 by Harry LewisYou can’t get the bits if you don’t have a pipe big enough to get them to you. Because of the engineering, political, and geographic challenges of the Internet pipe problem in the US, there has been a lot of attention focused on broadband diffusion — getting high speed connectivity to lots of people. Of course, businesses compete on their ability to deliver bandwidth — at least where there is competition at all. Mostly in the phone space, it seems, though the Comcast-Verizon FIOS argument certainly generates a lot of advertising revenue locally.
But there is another important Internet resource we almost never think about: IP addresses. Or IPv4 addresses to be precise, the 32-bit unique numbers that identify where the bits are supposed to go when any kind of message is sent to your computer. If you go to WhatIsMyIp or any of a number of other sites, you can see your own, shown as four numbers, each less than 256, separated by dots. That’s about 4 billion possibilities in all, a number that seemed unimaginably extravagant at a time that computers were huge.
Today, with computers in everything, even your wristwatch could use its own IP address, and plenty of devices smaller than that. The pool of IP addresses, which were divided into blocks and given out to nations, and within nations to companies and universities and governments, is being rapidly depleted. Not quite as rapidly as had been feared, but still: A story yesterday reported that there are only 625 days of IP addresses left, at the rate the reservoir is draining.
Various things can be done to stretch the supply. Some parties who got more addresses than they really need may be willing to surrender them voluntarily. NAT technology allows multiple connections to use a single IP address, but there are costs to that.
Sooner or later, we are going to run out of addresses. The solution is already known — IPv6, which uses 128-bit addresses. The code is already in the operating systems of computers being shipped today. But the switchover is likely to be hell — think of the switch of broadcast television to digital, with granny suddenly unable to get her soaps. Except that this switch will have a deadline attached to it.
624 days and counting, according to the current estimate …
January 13th, 2010 at 3:35 am
I think the switchover hell is greatly exaggerated.
To continue the TV analogy, it is important to remember that there were two very different transitions. Cable subscribers did nothing at all and continue to watch analog TVs: the cable companies were able to preserve the analog signal interface in the home even if the rest of the network had abandoned it. It was the rabbit ear antenna-using TV watchers who had to act, buying converters or new TVs, and unfortunately that segment of the population was the less tech savvy of the two.
From an end user point of view, the IPv4 -> IPv6 transition should be much more like the cable experience than the rabbit ear experience. The ISPs already use different technology to deliver the connection to your home (cable, DSL, satellite) than they do inside your home (ethernet, WiFi), so everyone already has a box to convert between these two formats. These boxes already NAT many reserved addresses onto the single external internet address assigned to your connection. If that external address has to switch to IPv6, the ISPs can reprogram the cable/DSL modems to do that without making you give up IPv4 on your home network. After that switch, whatismyip.com will say you’re using an IPv6 address even though your home computer still thinks you’re using IPv4, just as now it says your address is, say, 18.26.4.9 instead of 192.168.1.2. There’s no equivalent of rabbit ears here, no technology without a box in the middle. Even dialup users can be converted by putting the NAT box on the ISP’s side of the phone lines. Most ISPs already charge extra for public, fixed IP addresses that could be used to host a home server accessible from elsewhere. It’s a small step to giving the rest of the customers NAT’ed addresses.
The other half of the story is whether and when the big players at the core of the network will convert their networks. I don’t know what fraction already have, but there’s a strong financial incentive to be prepared to handle IPv6 and have it deployed before the IPv4 pool is empty, so as to avoid rejecting new customers. I trust this incentive to take care of the core network.
So with the edges taken care of by NATs and the core taken care of by not wanting to go out of business, I think ultimately, when it does happen, the IPv6 transition will much more like Y2K: a non-event thanks to advance planning and adequate preparation spurred by strong economic incentives.
January 13th, 2010 at 11:50 pm
Nice. Thanks, Russ. Let’s do it sooner rather than later, then. Bad enough we have wars about oil, let’s avoid having any about numbers!
March 14th, 2010 at 3:12 pm
I had been thinking should you had further links about the subject