You are viewing a read-only archive of the Blogs.Harvard network. Learn more.

Math is Hard! – Bad Journalism, IPv6, and the BBC

October 27th, 2003

Here’s a good way to frighten yourself: Learn about something, and then read what the press writes about it. It’s astonishing how often flatly untrue things get reported as facts.

I first observed this back in 1997 when I was a Democratic lawyer in the U.S. House of Representatives working on the (rather ridiculous) campaign finance investigation. (The investigating committee’s conspiracy-minded chairman was famous for shotgunning pumpkins in his backyard in order to figure out exactly how Hillary snuffed Vince Foster). The investigation was heavily covered by the press; as an investigator, I was privy to a lot of inside information, and got to watch how the underlying stories got reported. What was not surprising was how easy it was for White House officials and Congressional staffers to manipulate reporters by the use of leaks, exclusives, and off-the-record briefing. What was very surprising, though, was how lazy most Capitol Hill reporters seemed to be. Many simply did not undertake to check and verify the facts their insider sources fed them; and once something false was published as fact, most reporters were extremely reluctant to go back and correct the record. Most reporters preferred the pursuit of scoops and leaks to the gumshoe work of investigative reporting. (Some of the reporters I observed did excellent, factually rigorous work, but they were in the minority.)

More recently, I’ve seen the same discouraging phenomenon in reporting on technology and, in particular, the Internet.

Today’s Example A: a BBC story on the need for IPv6, by Ian Hardy. Let’s do a basic fact-check, line by line:

BBC ClickOnline’s Ian Hardy investigates what is going to happen when the number of net addresses – Internet Protocol numbers – runs out sometime in 2005.

The claim that IPv4 addresses are going to run out in 2005 is patently absurd. There is not a shred of evidence to support it.

Indeed, had the reporter bothered to contact the RIPE NCC or any of the other three regional address registries (the regional organizations responsible for assigning IP addresses to ISPs), or the IANA (the global organization responsible for allocating IP addresses to the regional registries), he would have been pointed to (gasp! ) publicly available data! For example, he might have looked at the October statistics presentation published jointly by the regional address registries. Or he might have Googled up Geoff Huston’s excellent July 2003 paper, “IPv4 – How Long Have We Got?“. I won’t repeat his analysis, but I recommend the paper — like most things geoffhustonian, it’s well-written, straightforward, and true to the data. Geoff’s notable contribution is that he uses BGP routing table data to supplement the address exhaustion pictures painted by the IANA and regional registry tables. With a bunch of careful caveats about how difficult and unreliable it is to predict future growth in demand, Geoff uses statistically reasonable projections to argue convincingly that the IPv4 space will likely last until around 2022. In any event, there is no evidence that IPv4 addresses will be exhausted in the coming decade.

For the BBC to report as a fact — in the boldface header, no less — that IPv4 addresses are going to run out in 2005 (i.e., within 2 years!) mirrors the atrocious quality of technology reporting worldwide.

Bad journalism?  Yes.  But wait! There’s more…

A taskforce of experts hope to solve the problem by creating what is called IPv6 and would provide 64 billion extra IP addresses.

IPv6 is already created; deployment started in 1999. And the sentence massively understates the size of the IPv6 address space. IPv6 replaces the 32-bit address field of IPv4 with a 128-bit address field. Doing the math, IPv4 has around 4.2×10**9, or 4.2 billion, unique addresses. IPv6 has around 3.4×10**38 (that’s 3,400,000,000,000,000,000,000,000,000,000,000,000,000), or 3.4 gazillion. A much, much, much bigger number than 64 billion.

Anyone who logs onto the internet will automatically receive an IP address.

True, sort of, but not every Internet device receives a publicly-routable IP address (which is what matters for address exhaustion analysis). Those who dial-up their ISPs or connect via corporate networks, for example, will receive non-public IP addresses on a temporary basis. It would be nice if every Internet-connected device had its own publicly-routable IP address, but that’s not the world we live in. Many end-users join the Internet from behind Network Address Translation (NAT) boxes. (NAT is a method of connecting multiple computers to the Internet using one publicly-routable IP address.)

So what’s wrong with this sentence? It shows that the reporter hasn’t bothered to learn the basics of IP addressing. The implication is that every new Internet device needs a IP address (and we all know about how many new Internet devices are getting bought up). In fact, the wide deployment of NAT (which is a bad thing, for reasons I’ll blog about later) allows huge numbers of new Internet devices to go online with only small numbers of IP addresses. You can’t speak intelligently about IPv4 address exhaustion without mentioning the impact of NAT. (The reporter later alludes obliquely to the difference between static and dynamic IP address, but without understanding it.)

The global distribution of available IP addresses is extremely unbalanced. Most of the numbers remain in the USA, where the technology was originally invented.

That’s at least a bit misleading. The global distribution of IP addresses mirrors the global distribution of the Internet, which is unbalanced. The policies of the regional addresses registries ensure that IP addresses are allocated and assigned to the networks that need them, worldwide.

More than two-thirds of the world’s IP addresses were bought by American companies.

Wrong. First off, IP addresses not “for sale.” Americans companies receive allocations and assignments as members of regional Internet registries, or as customers of members. IP addresses cannot be bought or sold. And take another look at the actual statistics. American companies have been assigned vastly less than two-thirds of the world’s IPv4 addresses. (It is true that the IANA assigned around 90 top-level IPv4 blocks to various companies and government agencies in the late 1980s and early 1990s, but those blocks were not “bought,” comprise less than a third of the total IPv4 space, include some non-US entities, and are vastly underused – it’s not like Apple Computer actually uses much of its top-level block.)

“Level Three Communications, which is a really large ISP, has more IP addresses than the whole of Asia,” said Matthew Sarrel, Technical Director of PC Magazine Labs.

Wrong again. The public data above shows that APNIC, the IP address registry for the Asia/Pacific region, has been allocated eleven (11) top-level blocks of IPv4 addresses, and three (3) sub-TLA blocks of IPv6 addresses. Level Three has been allocated zero (0) top-level blocks of IPv4 addresses (i.e., something considerably smaller than a top-level block) and zero sub-TLA blocks of IPv6 addresses. This information is not secret. It’s right here on the IANA website, with separate tables for the IPv4 and IPv6 spaces. All Google-able within a click or two.

“As companies and people in Asia get more devices they are going to run out of IP addresses.”

Not true. As note above, the IPv4 space is good for at least another decade, and probably two. The global IP addresses allocation system provides that those need IP addresses can get them, on the basis of need. The article somehow implies that those in Asia are at some kind of relative disadvantage, in terms of getting IP addresses. That’s just not true. The IANA policies for allocation of IPv4 address blocks to the regional Internet registries are applied evenhanded on the basis of need. The regional registries are all non-profit membership organizations dedicated to the service of their members, and to the achievement of IP address availability (as well as conservation and aggregation). My guess is that the reporter did not bother to talk with anyone at one of the regional Internet registries. In any event, he didn’t fact-check the quote above.

One of the biggest pressures on IPv4 is the ‘always on’ internet connection. At the moment, when you dial your ISP they assign you a temporary IP address, which is taken away the moment you log off and given to someone else.

Not so. Always-on Internet devices can be configured either with private addresses or publicly-routable addresses. ISPs decide whether or not to assign their customers publicly-routable addresses or private addresses behind an NAT box, taking account of a variety of network management considerations. Whether customers’ devices are always-on or not may well affect those ISP considerations, but it’s not correct to state that always-on devices automatically require publicly-routable, static IP addresses.

But in the new era of 3G wireless computing, each of us needs a static, or permanent IP address.

That suggests that static IP addresses are (or should be) assigned to individuals, one address to one person. Wrong: IP addresses are assigned to Internet devices, by the relevant service providers. In the future, we will want static IP addresses for each Internet-connected device we have.

This is intended to provide four billion times four billion times four billion as many as currently exist.

Aha! Now we learn why the reporter misreported the total number of IPv6 addresses: He can’t multiply. “Four billion times four billion times four billion” isn’t the same as 4 x 4 x 4 = 64, and then you just tack the “billion” back on, giving you the story’s “64 billion” figure. (You have to multiply out all the zeros, too).


So: Do these factual errors matter? (I.e., aren’t I being a little hard on the poor reporter? After all, he’s a journalist, not a Internet techie, and he’s got a lot of stories to write.)


  1. It matters because the story paints a false and alarming picture. “Eeek! IPv4 addresses are going to run out ‘sometime in 2005’! Asia is being treated unfairly by the Americans! We’ve got to do something!” Readers reasonably expect the BBC to report reality; this article amounts to scare-mongering.
  2. It matters because there is a real and important IPv6 story that should be reported. Even though there’s no crisis in IP addressing, there is nevertheless an interesting and important story to be told. IPv6 is needed; it’s a powerful upgrade to the Internet’s core protocols; it will enable new capabilities and possibilities by re-enabling a more purely end-to-end architecture for the Internet. There are serious technology and policy implications to the question of whether the Internet is truly end-to-end or not. The false alarmism of the BBC story obscures understanding of the real issues at stake.
  3. It matters because it’s the BBC. The BBC is a news agency of global scope, and a reputation for reliability. It should be able to report on technology with some degree of accuracy. I expect more from the BBC.
  4. It matters because it was easily avoidable. All of the factual errors in the story could have been corrected with some decent reporting. For example, none of the individuals quoted work for one of the regional IP address registries, which are the obvious sources of expert factual information about the resources they are charged with administering.
  5. It matters because of what the error rate implies more generally. IP addressing is something I know fairly well. If the BBC is making such fundamental misstatements of fact about something I know, it is reasonable (and troubling) to wonder whether similar ratios of mistake to fact arise in the rest of the BBC’s reporting.

ICANN Non Delenda Est

October 27th, 2003

The latest news from the ICANN meeting in Carthage is that substantive issues, rather than organizational hoopla, are dominating the agenda and the discussions.  That’s a welcome story, and indicates the success of last year’s ICANN reform process.  Key issues:  DNS wildcards, IPv6, Whois, security, new top-level domains. 

This is the first ICANN meeting I’ve ever missed, and it feels weird to be so far from the action.  I’ll be following via webcast later this week.  A big shout out to the ICANN crew:  Boo-ya!  It’s great to see the organization in such solid shape.  The new leadership team is settling in quickly, striking out in new and useful directions, and doing an excellent job so far.  Notably, CEO Paul Twomey has devoted considerable resources to increasing participation from developing countries.  And indeed, I’m happy to note that the Carthage schedule is full of workshops and more workshops and regional meetings and more regional meetings at least partially focused on the imperative of internationalization in policymaking for the Internet’s systems of unique identifiers.