[I’m re-posting this, because the original, from Saturday, seems to have disappeared.]
The big Internet tech news last week was that the last block of Internet addresses, for the version of the Internet Protocol (IP) that we mostly use (IPv4), has been allocated. Or, as the headlines said, we have now run out of Internet addresses. Of course, it’s filled the tech media, as above, but it’s shown up in the mainstream press as well; here it is from the New York Times, and from The Guardian.
What does it really mean, that we’ve run out of IPv4 addresses?
Well, for one thing, it doesn’t mean that we’ve run out of IPv4 addresses. The Times gets it better than the other articles, in its headline:
The Last Block of IPv4 Addresses Allocated
The last address has not been assigned, not by a long shot. IPv4 addresses are allocated to organizations in large blocks — sometimes blocks of 60,000 or so, sometimes blocks of more than 16 million. Those organizations then assign addresses within those blocks, sometimes individually and sometimes in sub-blocks. What has just happened is that the last large block of addresses has been allocated. There are still many, many IPv4 addresses available for assignment, within many of the blocks that have been allocated.
For example, IBM has a 16-million-plus block of addresses comprising all addresses that start with “9” (that is, every address of the form 9.x.x.x; they also have some of the 129.x.x.x range). Those 9.x.x.x addresses are assigned within the company’s network. Not all of them are assigned, of course; there aren’t more than 16 million devices within the company.
Similarly, Internet service providers, such as Comcast and Verizon, have large blocks of their own, some for use within the company, and some to provide to their customers.
Many companies have blocks that are much larger than they need, far more than they could ever imagine using for their normal networks. Those blocks were allocated to them in earlier times, before the worldwide web and the explosion of Internet usage, when we never thought it would matter. Or they were assigned later, when we assumed that IPv6, with many orders of magnitude more addresses, would be well deployed by now. (I’ll note that it would be very difficult, even though large portions of the allocated blocks remain unused, to reclaim the unused bits and to reallocate them.)
Let’s not be Chicken Little, here; the sky is not falling, an the Internet is not imminently doomed. Indeed, the Internet will mostly run fine, as it is, for many years yet. We’ll all be able to read our email, buy from Amazon and eBay, use Facebook, and see YouTube videos.
Eventually, we’ll be crowded out by expanding Internet use, though we have techniques to keep that at bay for a long time. What will be blocked by this are — and this should be a familiar refrain to readers here — new applications, new uses of the Internet. To move into the future, beyond email and eBay, Facebook and YouTube, we need to move to IPv6.
We have enough IPv4 addresses for now, and for a while, to accommodate putting every computer on the Internet, as long as we’re thinking of “computer” as we have been: desktops and laptops. Maybe iPads, too. But now add Kindles and other eBook readers. Add smart-phones. Consider that every mobile phone is a smart-phone. Do we have enough v4 addresses for all of that?
Now move into the Internet of things: add every car, because our cars need to be online. Add every television (they’ll stream video directly), every stereo receiver (streaming music, “radio stations”, and other audio from the Internet), every portable music player from “boom box” to iPod Nano. Are we getting there? Include appliances: alarm clocks, refrigerators, coffee makers. Include home- and building-automation targets: thermostats, light switches, and so on. Put in sensor networks, traffic-control and monitoring systems....
Well, given all that, we ran out of v4 addresses long ago. It’s not really the v4 address-space depletion that should be driving the move to IPv6, but the need for more address space for future applications. If you don’t think that sort of thing is important, consider this news item about electric-grid problems stemming from the recent ice storm in Texas:
FORT WORTH, Texas — A high power demand in the wake of a massive ice storm caused rolling outages for more than eight hours Wednesday across most of Texas, resulting in signal-less intersections, coffee houses with no morning java and some people stuck in elevators.
The temporary outages started about 5:30 a.m. and ended in the afternoon, but “there is a strong possibility that they will be required again this evening or tomorrow, depending on how quickly the disabled generation units can be returned to service,” the chief operator of Texas’ power grid said in a release.
Consider the potential consequences of intersections without traffic signals and people stuck in elevators. We’d like to shut the power down in an area selectively, killing most of it but leaving the elevators running (at least until they open on the next floor), leaving a trickle of emergency lighting, leaving the the traffic lights running. We can do that, if everything’s addressable, and the power control system is set up to allow distribution with sufficient granularity.
But if it takes a “Chicken Little” scare — “The Internet is falling! The Internet is falling!” — to get IPv6 out there, well, here it comes.