The Internet Uncertainty Principle

Put aside all of those internet research reports for a moment and think about the data we'd like to have but cannot obtain. The absence of data creates uninformed policy, but common sense dictates an obvious solution to broadband problems as the United States falls farther and farther behind.

by Alex Goldman ISP-Planet Managing Editor [March 19, 2007]

KC Claffy of CAIDA, the Cooperative Association for Internet Data Analysis, spoke at the Freedom to Connect conference about the limits of our knowledge about the internet.

One of CAIDA's most accessible projects (and the reason I'd heard about them) is CAIDA's maps of the internet.

Claffy's spent her career on higher level theory, starting with a dissertation titled "internet traffic characterization: a methodology to support more accurate workload characterization in the face of increasing diversity in internet traffic types and qualities".

At one point, she decided to go directly to engineers and ask them about internet issues. "I came up with 100 problems. I divided them into 16 taxonomies. But eventually I realized that every one of them was about ownership or trust."

Heisenberg's internet uncertainty
She has a list of things we cannot measure: "the topology of the internet from one point to another (we cannot measure a path), we cannot get a router to give us its entire state (including its next best routes), we cannot get a precise one way delay, and we cannot obtain an hour's worth of packets in the core. I cannot tell you how much spam, phishing, viruses, or infected hosts are on the internet."

"There's no 'why can't I reach google tool' (although my mom thinks I should have one). I cannot measure your throughput (although you can measure your own throughput to a specific point with a speed test)."

She said that at the recent FTC workshop, the government seemed to think that "a tool" would enable it to measure net neutrality and determine whether or not net neutrality was being violated. "There are no tools," said Claffy.

This lack of tools has a serious consequence. "We cannot weed out false beliefs," said Claffy. False beliefs, such as the rate at which the internet is doubling, have led to massive misallocation of investment and contributed significantly to the telecom recession (which may or may not be over, depending on whom you ask).

Claffy said that if you look at the telephone network, you'll see a tremendous amount of resources devoted to measurement and billing (ISPs reply that these resources are devoted to double billing). Claffy says that building the internet without measurement systems lowers costs, but at a hidden price. In the future, she says, we may need to charge per packet.

The internet's a private toll road
ISP-Planet has argued in an editorial that we need usage-based pricing but there is strong opposition to it. Many today associate usage-based pricing with third world dialup.

Brad Templeton writes that usage-based pricing would kill innovation:

I think a move to usage-based pricing would be tragic. As I have written before, usage-based pricing requires that every new app that makes noticeable use of bandwidth has to be financially justifiable (rather than simply interesting) from the get go. All sorts of new, experimental apps that surfed on the spare bandwidth would never have been developed in a usage-based pricing world, and many new undreamed of apps of the future will never be developed if we go down that road.

Claffy would prefer to see the backbone described by Yochai Benkler in Property, Commons, and the First Amendment: Towards a Core Common Infrastructure (.pdf).

Benkler writes:

Just as public highways and sidewalks enable anyone to walk as and when they please, so too the core common infrastructure will be open to all. Having public highways does not eliminate commercial provision. Alongside highways we see railroads and airplanes providing private proprietary services with advantages and disadvantages relative to the open road.

On the highways themselves, we see all kinds of uses, interrelated and coexisting. Easy Rider and Thelma & Louise use the same highway system that Ryder rent a truck and Federal Express do. So too, the common core infrastructure need not be built as a nature preserve for nonprofit users, but as a facility for all. The primary difference between the common infrastructure and the proprietary infrastructure is that no one will have an exclusive right to capture and control the social benefits of its use.

Of course, a public infrastructure could be measured, if it were so designed. But those arguments lie in the future. ISP-Planet doubts that the United States will build a public information superhighway, but any of the world's top 10 broadband nations might.

Perhaps local governments across the U.S. will build local internet roads, and once those roads are built, the government will help connect them.

— End