Vulgar Net Neutrality

Introduction #

This is a very simple story: A government agency tried to make a law, failed, and is now trying again. The end.

If that sounds too boring, you can also tell it like this: A government agency is trying to destroy the Internet, and here is what you must do to stop it RIGHT NOW!!!

I read that story in 2014. I’m an engineer at one of those big Internet companies, so naturally I was interested: if it’s true, I might soon be out of a job. If you are reading this in 2014, you must have heard the story as well. It has been in the headlines, and is pretty exciting. It has villains, heroes, money, power, the entire planet, an apocalypse, it’s actually pretty easy to understand, and… Wait, I forgot.

You see, if you’re a computer engineer, it isn’t any of those things. I heard the story over and over, and I thought to myself: “That just doesn’t make any technical sense; but this story is everywhere, and no-one is saying anything, so what am I missing?”

The story was very hot. Articles about it in May, 2014 had headlines like these: “Outside the FCC and Nationwide, a Single Call: ‘Save the Internet!”¹ “Nationwide protest at FCC Headquarters to 'Save Internet’”² “FCC Votes to ‘Fast Track’ Death of Net Neutrality.”³ “Your complete guide to the murder of net neutrality.”⁴ “FCC Proves Yet Again That It’s Out to Kill Net Neutrality.”⁵ And then there were all the hashtags: #SaveTheInternet, #netneutrality, #StopTheFCC, #thiscannothappen, to name only a few.

Is it really true? Is the Internet about to be destroyed? Is there anything we should do to save it? What’s been going on?

If we try to answer these questions by reading news articles and blog posts, we find that virtually all the opinions fall into one of two camps. One camp is for net neutrality, and the other is not. According to those in favor of it, the Internet is about to be destroyed by so-called “fast lanes”, in which web sites pay extra to make their sites load faster than the norm. A government agency, the Federal Communications Commission or FCC, has the power to outlaw fast lanes, but is proposing to legalize them instead. Its critics say that small start-up websites will never be able to afford these fast lanes, and so Internet innovation will dry up and disappear. The colorful and ever-changing bazaar of the Internet will be replaced by the equivalents of Burger King, Disney, and Starbucks.

According to those opposed to net neutrality, the fast lanes are simply a way for bandwidth hogs like Netflix to pay for improvements to the Internet that they are clogging. In this view, successful Internet businesses, whose surging traffic puts a severe burden on the Internet’s infrastructure, ought to pay for its upkeep instead of getting a free ride like they do today.

Both these views seem reasonable. Both are irreconcilable. But the real problem is that they are both hopelessly wrong.

Little of this came about intentionally. Many of the facts surrounding net neutrality are corporate secrets, revealed, if ever, to only a few journalists and academics who cannot share them, and who have had to earn the trust of their sources over many years. Thus even the most expert analyses are based on supposition, inference, and guesswork, as well as engineering and legal intricacies that are difficult to explain. Combine this with an intense pressure to sensationalize the news, and even the most intellectually honest reporters and editors will be disinclined to question net neutrality stories, if they are easy to understand and grab the headlines. If the stories include an element of moral outrage, political activist groups will disseminate them further, in the never-ending quest for donations and voters. This fuels more news coverage, which fuels more activism, until there is a raging and bitterly divisive political blaze that can take years to burn itself out.

This is what has happened to net neutrality. It has become as polarized an issue as gun control or gay marriage. The dominant versions all inspire moral indignation, and are also incoherent in ways only experienced engineers might easily notice. I refer to all these views as “vulgar net neutrality”.

In what follows, we will first debunk the two most prominent vulgar net neutrality and vulgar anti-net neutrality views. Then we will look at how the meaning of net neutrality has evolved over many years, why it matters, and what the Federal Communications Commission (FCC) has to do with it. We will see what options the FCC has had to choose from, and how vulgar net neutrality has distorted them; and finally, we will conclude with some thoughts on what would now be best for the Internet.

This is a story with many reversals and surprises, yet in the end it remains quite simple: a government agency tried to make a law, failed, and is now trying again. But it is also a story about how little we know about the Internet, and of how much we stand to lose if we do not learn in time. This is not a story about good and evil, but one that is important for a more modern reason.

Because in this story, the devil isn’t in the detail. It is in the engineering.

Table of Contents #

Introduction #

Fast lanes I have known #

The woes of giant corporations #

What is net neutrality? #

Why does net neutrality matter? #

What is the FCC? #

Net neutrality becomes law #

Verizon v Fcc #

Common carriers #

“Commercially” #

The FCC’s options #

Option 1: Do nothing #

Option 2: Add “commercially” to the Open Internet Order #

Options 3 & 4: The Title II and Title II “lite” choices #

Four lessons #

Conclusion #

Fast Lanes I Have Known #

I am sitting on a small island in Indonesia. It is mid-morning in May, 2014, and very bright. Along the horizon huge cumulus clouds billow upwards for thousands of feet, like gigantic Himalayan peaks, their crests a dazzling white above the thunderous grays of not-so-distant rain storms. Buried among the tropical trees behind me are scores of small dilapidated houses, with the rusty corrugated metal roofs one sees everywhere in Sulawesi. In front of me is a small laptop. A browser is open, and it is slowly loading dozens of pages in different tabs.

Back in Washington, DC, demonstrators are camped outside the FCC in tents, and are about to disrupt its televised hearings. The articles I’m loading attack the FCC as well. They all make the same point: the FCC is going to destroy the Internet by allowing big corporations to pay for “fast lanes” that will load their web sites faster on everybody’s phones and computers than the rest of the Internet.

The wifi here in Sulawesi is terribly slow. Whenever a distant web site sends me a page, a quarter or more of it is lost - the technical term is “dropped” - and my browser has to re-request those bits of the page, again and again and again. The browser just gives up on the images, showing “broken image” icons instead. The pages load excruciatingly slowly, taking minutes at a time.

But not all of them. If I point my browser at http://cnn.com, based in Atlanta, Georgia, that page loads in just 20 seconds. If I want to, which I don’t, I can instantly read the drivel that CNN considers headline news: Magic Johnson’s response to Donald Sterling’s racist rants, Mariah Carey’s struggles to fit in somewhere, and fresh video of a husband in Pakistan just after his wife was stoned to death. On the other hand, the page of my American bank ultimately never loads, despite repeated attempts. Why is this? Is it because CNN has more money than my bank, and can afford better and faster servers? Impossible: my bank’s market capitalization would allow it to buy CNN many times over.

No: The reason CNN and some other large corporations’ sites load faster in Indonesia and the rest of the world, is that they have paid to use Internet fast lanes. Web sites have been paying to use fast lanes for almost sixteen years, fast lanes that the demonstrators outside the FCC think do not yet exist. Yet the sky hasn’t fallen, the Internet hasn’t been destroyed; the Armageddon that the net neutrality FCC demonstrators fear never occurred.

Quite the contrary. Google was founded in 1998. Wordpress and Skype were created in 2003. Facebook went live 2004, Reddit and YouTube in 2005, Twitter in 2006; the list goes on and on. Every one of these companies began life as a tiny start-up, sometimes in somebody’s garage, and they all succeeded in spite of the Internet fast lanes available to their competition.

Of course, this doesn’t mean that these fast lanes led to the success of the Internet; but it does mean they didn’t stop it.

However, we must be careful not to overstate the case. First of all, net neutrality experts aren’t worried about this kind of fast lane, but a different one we will look at later; and what worries them isn’t speed, but a lack of competition. Second, these lanes are intended to be “just-as-fast lanes”, not faster than everything else. CNN just wants to be as fast as the average website, in average places. It is only where Internet access is poor, as in much of the developing world, that they become “fast lanes”.

That said, there is a widespread view that any difference in delivery quality and speed that can be bought is a “fast lane” that will devastate competition and innovation on the Internet, and that support for it is morally outrageous. This is a vulgar net neutrality view, and the past sixteen years of just such fast lanes prove unequivocally that it is false.

The Woes of Giant Corporations #

Every year, every public corporation in America must complete and file Form 10K with the Securities and Exchange Commission; these forms are all publicly available on the Internet.

Should a company’s stock price seriously decline later that year, and the company could have reasonably foreseen why but failed to say so in either Item 1A or Item 7 of Form 10K, the consequences are devastating. In addition to SEC fines, US law entitles shareholders to recoup their losses from the company’s remaining wealth and assets.

Let’s assume that Comcast, Verizon, AT&T and other giant Internet service providers each believes that its network has been swamped with traffic from the likes of Netflix; that upgrading a network to handle this traffic would be exorbitantly expensive, perhaps so expensive that Netflix ought to pay for part or all of it.

If this is the case, we should expect to read all about it in at least one of the 10K forms these companies have filed in the past several years.

Instead, there is nothing.

What we find are worries about revenue lost as more customers cancel their telephone landlines forever; about revenue lost to customers who “cut the cord” of cable TV in favor of entertainment on the Internet; and, before 2014, of uncertainty over the outcome of network regulation working its way through the courts. In fact, rather than invest in new or improved broadband Internet networks, these companies appear focussed on signing up more customers for their existing networks.

So the view that Netflix and similar companies are putting giant Internet service providers like Comcast and Verizon in distress is implausible, because there is nothing about it in the 10K forms. Like other vulgar net neutrality views, this vulgar anti-net neutrality view inspires moral indignation, but is at odds with the facts.

What is Net Neutrality? #

There is at least one thing everyone writing about net neutrality agrees on: the term was first coined in 2003 by Tim Wu, then a law professor at the University of Virginia, in an academic paper titled “Net Neutrality, Bandwidth Discrimination”.⁶,⁷ Tim Wu maintains a net neutrality web page now; the first sentence there is: “Trying to figure out the network neutrality debate via the web is kind of hard.”

He’s right. Everyone who sets out to explain what net neutrality is always seems to end up writing the same paragraph. In it, they first they say Wu coined the term in his 2003 paper. Next, they say the consensus definition of net neutrality is the one Wu described. Then there is a sentence that starts something like: “Net neutrality is,” immediately followed by the inevitable qualifier: “essentially”, “basically”, “briefly”, “something like”, “more or less”, “roughly”, etc. The sentence then continues with a somewhat woolly description of what net neutrality “essentially/roughly/etc” is, and while the descriptions are sometimes quite different, they all agree that Wu’s net neutrality is about treating all Internet data equally, or “neutrally”.

So let’s not try to figure out what net neutrality is from secondary sources. Instead, let’s go back to the source: the 2003 paper by Tim Wu.

The idea of net neutrality #

One of the first points Tim Wu makes is that net neutrality has historically been voluntary, a sort of “self-regulation”: over the decades, there has been a sort of gentleman’s agreement on the Internet to act net neutrally. It has been a bit like a moral principle, like “thou shalt not steal”, and nobody can really say why it has worked so far.

A second important point about net neutrality is that it is an idea like justice, poverty, love, and so on. We know, or we think we know, what each of these things are, even if we cannot define them. We know it in our gut. Despite millennia of attempts, no one has found a satisfactory definition of what love is. Great writers may be able to capture it, but that is not the same thing as being able to explain what love is, especially to someone who has never experienced any form of it. Justice may seem as difficult as love, but in fact there have been many classic attempts to define justice, and some of them have found their way into countries’ constitutions.⁸

Net neutrality is a concept that falls somewhere on this spectrum. This hasn’t stopped people from trying to define it though, or from trying to say what a net neutrality policy should be.

The two are more tightly linked than we might think. Consider poverty, and what a country’s poverty policies ought to be. Let’s just call the country Placeonia. In Placeonia, the government defines poverty as a daily income below one euro (or dollar, or whatever), researches the causes of poverty, and enacts policies to address them. Then one day, someone discovers that in Placeonia, on average, people are only “poor” for two years out of ten. Everyone just assumed poverty was a permanent trap, but in Placeonia, that turns out not to be true. What “poverty” actually means, how it should be defined, and what poverty policy should be, all subtly change in Placeonia.

In his paper, Wu described two versions of net neutrality and its policy that had emerged in the preceding years, and then proposed a third version he thought was better. Wu was under no illusion that his version would survive the test of time:

“Neutrality, as a concept, is finicky, and depends entirely on what set of subjects you choose to be neutral among. A policy that appears neutral in a certain time period, like “all men may vote”, may lose its neutrality in a later time period [e.g. when women demand that right; rbondi]… ” ⁹

Wu pointed out that this had already happened with network neutrality. To see how, we need to know more about how the Internet works.

Net Neutrality as Packet Neutrality #

Computers send information to each other as pulses. Over a wire they are pulses of electricity, over a fiber optic cable pulses of light, and over wifi and radio, pulses of electromagnetic waves. The pulses are extremely rapid, fired faster than any machine gun fires bullets, and each contains very little information. A weak pulse is interpreted as the number zero, and a strong pulse as the number 1. That’s it: computers just send each other streams of 0s and 1s over a network.

A computer that receives a stream of 0s and 1s faces the same problem we would if we received a stream of letters with no spaces between them. To make sense of the letters, we would need to “repackage” them into words using spaces, into sentences using punctuation, and so on. To oversimplify, the Internet groups 0s and 1s into units called packets. Packets travel over the Internet like cars on busy highways.

When the first Internet was designed and built in the 1970s, one of its guiding principles was that it would be neutral about packets. As a packet travelled across the network from computer to computer, each would look at the packet’s delivery address and ask: “Which computer should I send this packet to next, for it to arrive most quickly at its destination?” The answer to that question is never more than an educated guess, even today. The Internet was designed to be “packet neutral” in the sense that whether a packet was part of an email from the President of the United States, or an email from me, it would get the same treatment.

In the Internet’s earliest years, packet neutrality worked well. Nobody cared too much if their emails arrived a minute earlier or later, not even Presidents. Nobody even dreamed of video over the Internet. Today, however, we care a great deal whether a packet that is part of the movie we are watching arrives a minute earlier or later. But the Internet treats a movie packet the same way it treats an email packet. That’s a kind of neutrality nobody really wants. In fact, it isn’t even neutral: email packets are getting preferential treatment, in the sense that all packets are treated in a way that is fine for email packets, but not for movie packets.¹⁰ As Tim Wu said, “A policy that appears neutral in a certain time period, like ‘all men may vote’, may lose its neutrality in a later time period …”

Net Neutrality as “Open Access” #

In 2003, when Tim Wu published his famous paper, the net neutrality debate had changed. It was about the rise of cable companies as Internet service providers, or ISPs – companies that connect home users like you and I to the Internet. Until then, ISPs typically had been small local companies; now giant cable TV providers like Comcast and Verizon offered home Internet access as well.

Net neutrality advocates were afraid that cable companies would start “bundling” Internet access the way they bundle cable TV. You can purchase basic cable TV, or you can pay for ever more expensive premium “bundles” that include popular channels like HBO, ESPN, and so on. The worry in 2003 was that ISPs would start to sell a “basic” Internet service with, say, only email and access to a handful of websites, and charge more for “bundles” of access to instant messaging, audio, video, the most popular websites, and so on. This kind of “bundling” violates the general principle of net neutrality, because it discriminates between kinds of Internet data.

It never happened – ISPs don’t bundle access in this way – but of course, ten years ago, nobody could know that it wouldn’t happen. At the time, people wanted to fight Internet bundling, but they couldn’t use the “packet neutral” version of net neutrality against it. Bundling didn’t stop packets ranging wide and far, and neutrally, over the Internet; bundling kicked in only at the fringe of the Internet, when the packets spilled onto a home computer.

So net neutrality advocates came up with a different principle, called “open access”. It’s a little difficult to describe the specific problem “open access” was designed to solve, because technology has changed so much in the ten years since. Today on our desktop computers, we use only one kind of program, a web browser like Chrome or Internet Explorer, to do everything: watch movies, listen to radio, send instant messages, read email, browse the web. Ten years ago that wasn’t the case: you needed a different program to do each of these things. For example, most email could only be read with a program like Eudora or Outlook, not with a web browser. The principle of “open access” simply stated that an ISP had to allow all programs to access the Internet, as opposed to “bundling” access by programs.

The trade-off problem and net neutrality #

Tim Wu’s 2003 paper had a very simple, straightforward goal: to argue that “open access” didn’t capture everything that was important about net neutrality, and to propose a better principle that did. Wu listed several problems with “open access”, but the main one was similar to “packet neutrality”. Even if “open access” forced an ISP to allow Internet video viewing programs to connect to the Internet without bundling, video packets were still treated just like email packets, resulting in bad video and good email service.

What “open access” was missing, Wu argued, was that “network design is an exercise in trade-offs… ” On an ISP’s network, the trade-off is between quality of service and treating packets equally. If we want our ISP to deliver our movies as smoothly as our emails, Wu argued, we have to let it treat the packets differently, to let it favor movie packets over email packets – to violate net neutrality. On the other hand, if we want our ISP to treat all packets equally, we have to sacrifice the quality of our Internet service: email will work well, but movies won’t. Wu argued that we can’t have both, because that would require “an impracticable upgrade of the entire [ISP’s] network.” All ISPs and the rest of the Internet could, in principle, upgrade their networks to handle both movies and email with ease, but only in the sense the United States can put another man on the moon in five years if it really wants to. In principle, this is possible; in practice, it would be ruinously expensive.

This issue, Wu said, was a matter of net neutrality, and it was critical; if no principle of net neutrality could be found to resolve it, the Internet might suffer. And so, in his 2003 paper, he proposed one.

Tim Wu’s Principle of Net Neutrality #

For Tim Wu, the new problem of net neutrality in 2003 was a problem of trade-offs. Either we could have good video and email, or we could have ISPs treat all packets neutrally; we couldn’t have both.

Wu proposed that we think of this as a discrimination problem, like sex or racial discrimination: the problem is whether to allow ISPs to discriminate between packets. Wu’s idea was that if we do that, we can then look to the law for inspiration about how to deal with discrimination, because it has been a legal issue not just for decades, but centuries. In the law, we find that there are two kinds of discrimination. One kind is clearly “troublesome” and forbidden, such as employers choosing whom to hire based on their religion, gender, race, or politics. Another kind, however, is clearly “permissible” and justified: it’s acceptable for an employer to choose between applicants based on their “education-level, intelligence, and demeanor.”¹¹

Wu’s idea was that we can make the same distinction between permissible and forbidden discrimination of Internet packets. He proposed that it is permissible for an ISP to discriminate between packets in order to improvequality of service, protect the network, and certain other fairly obvious reasons (Wu provided a list); but that discriminating for any other reasons should be forbidden. Furthermore, ISPs would only be allowed to play favorites with packets once they were inside an ISP’s own network:

“The effort is to strike a balance: to forbid [ISPs], absent a showing of harm, from restricting what users do with their Internet connection, while giving the [ISP] general freedom to manage bandwidth consumption and other matters of local concern. [My network neutrality] principle achieves this by adopting the basic principle that [ISPs] should have full freedom to “police what they own” (the local network) while restrictions based on inter-network indicia should be viewed with suspicion.”¹²

In other words, Tim Wu’s definition explicitly favors discrimination of packets by Internet service providers.

Let me repeat that. The definition of net neutrality that journalists, bloggers, and activists all invoke as the gold standard that forbids discrimination, Tim Wu’s definition from 2003, in fact emphatically endorses it. Whatever the net neutrality uproar of 2014 is about, it is definitely not what almost everyone says it is about: Tim Wu’s version of net neutrality.

This is one crazy story. Fast lanes don’t kill the Internet, ISPs are not worried about upgrading their networks, and Tim Wu’s net neutrality principle that is said to forbid packet discrimination, in fact supports it. If the news stories about net neutrality can’t even get these fundamental facts straight, perhaps we ought to just give up trying to understand it, on grounds of sheer incomprehensibility.

Which raises the question: Why should we even care about net neutrality?

Why does Net Neutrality Matter? #

We have seen that debates over what net neutrality is are about as old as the Internet itself. The net neutrality furor of 2014 claims to be about Tim Wu’s version of it, yet clearly isn’t. But why care about net neutrality at all? Tim Wu asked that question himself in his seminal paper, and the answer he gave in 2003 is still accepted today:

“So what is attractive about a neutral network — that is, an Internet that does not favor one application (say, the world wide web), over others (say, email)? Who cares if the Internet is better for some things than others? The argument for network neutrality must be understood as a … belief about innovation, one that has gained significant popularity over the last two decades. … A communications network like the Internet can be seen as a platform for a competition among application developers. Email, the web, and streaming applications are in a battle for the attention and interest of end-users. It is therefore important that the platform be neutral to ensure the competition remains meritocratic.”¹³

In 2003, adherents of net neutrality believed that it alone would ensure the best possible Internet hardware and software. They didn’t know yet what that would be; for example, the iPhone had not been invented, let alone the almost billion dollar phone app industry.

By 2010, awed by the Internet’s stunning successes, the Federal Communications Commission and many other people went further: they now believed that the Internet “has thrived because of its freedom and openness—the absence of any gatekeeper blocking lawful uses of the network or picking winners and losers online … ”¹⁴ – in other words, because of net neutrality.¹⁵

Today this is a mainstream view: even so-called opponents of net neutrality believe some version of it.¹⁶ What people differ about in 2014 is first, whether net neutrality ought to be the law instead of voluntary; and, second, whether a weaker version of net neutrality — for example, one that allows so-called “fast lanes” — will still permit the Internet to thrive.

It is important to keep in mind here that net neutrality is not an end in itself. Net neutrality is seen only as a necessary¹⁷ means to an end, and that end is the continued flourishing of the Internet and Internet innovation. That, and that alone, is why net neutrality matters.

What is the FCC? #

Remember, this is a simple story: a government agency tried to make something the law, failed, and is now trying again. By now you will have guessed that the “something” is net neutrality. The agency is the Federal Communications Commission, or FCC, and it began with radio.

Radio is just physics. Anyone can build or buy a radio and try to start their own radio station, broadcasting music and speech. In the late 1920s, that’s exactly what happened; but would-be broadcasters transmitted on only slightly different wavelengths, so that “stations were soon wildly blanketing each other and distracted listeners were assailed with scrambled programs.”¹⁸ You couldn’t tune into just one station; on almost every frequency, you would hear two or more stations talking over each other. The tragedy of the commons threatened to destroy radio.

Ironically for free market proponents, it was federal regulation that created the advertising-driving radio industry we’re all familiar with today. In 1926, only 4.3% of radio stations were “commercial broadcasters.” The rest were colleges, unions, and various civic organizations. By 1933, thanks to federal regulation, 97% of all nighttime broadcasting belonged to NBC and CBS, funded completely by radio ads, and most daytime transmissions had also become commercial.¹⁹

Flush with this regulatory triumph, in 1934 the United States Congress passed the Communications Act “…so as to make available, so far as possible, to all the people of the United States a rapid, efficient, nationwide, and worldwide wire and radio communication service with adequate facilities at reasonable charges…” ²⁰ Congress didn’t believe the hidden hand of the free market could accomplish this alone. Regulation would be necessary, and so the Act of 1934 also created the Federal Communications Commission, or FCC, to carry it out. It was to be led by two Democrat and two Republican appointees approved by Congress, with a tie-breaking chairman from whichever party held the US Presidency.

Then Congress did a slow, sixty-year somersault. The Telecommunications Act of 1996 overhauled the 1934 Act to remove as much regulation as possible. In the FCC’s words, the Act’s purpose was “to let any communications business compete in any market against any other.”²¹ The technology breakthroughs of the 1970s and 1980s, including cable TV and the Internet, had convinced Congress that the best way to spur innovation was to unshackle the market. But deregulation was just a means, not an end. The FCC’s mission remained what it had been since 1934: to foster world-class, worldwide, affordable and accessible telecommunications for the American people. In almost no time at all, that came to mean the Internet.

And now the FCC was faced with an existential dilemma. This thing, the Internet, kept changing; it simply could not be pinned down. Practically every month new and unexpected ways to use it were invented, even entirely new industries: Amazon, Facebook, Google, wifi, Netflix, the smart phone, tablets, and on and on. So just how affordable and accessible did the Internet have to be - what exactly was the FCC supposed to be enforcing? What exactly was the “best” possible Internet, and what was the best way to keep it getting better?

Did the Internet even need the FCC?

Net Neutrality becomes Law #

Throughout the first decade of the twenty-first century, the FCC tried to define what role, if any, it should play in the Internet revolution.

The Telecommunications Act just says that the FCC should ensure that there are world-class, “worldwide wire and radio communications service[s]” that are accessible and affordable to all Americans. The Act doesn’t say how the FCC should accomplish this. Instead, it gives the FCC two powers. One power is the legal right to determine that someone is violating the Act, threatening these services, and to fine and stop them. The other power is that the FCC may pass regulations that are more specific than the Act, and which are designed to enforce it. The regulations are interpretations of the Telecommunications Act; they are effectively laws, even though Congress did not pass them.

Companies and individuals can challenge FCC rulings and regulations in court, claiming either that the FCC exceeded its legal authority under the Telecommunications Act, or that the regulations are unconstitutional. Furthermore, the FCC cannot make or change any regulations without first proposing them to the public, and requesting their comments. They must do this by publishing the proposed regulations in a document called a “Notice of Proposed Rulemaking”.

The first public attempt to define the FCC’s Internet regulatory role occurred in 2004, when a Republican FCC Chairman, Michael Powell, gave a speech in which he laid down four “guiding principles” for preserving Internet freedom.²² These were not regulations, but they put American companies on notice about what kinds of actions the FCC might sue or fine them for.

In 2010, after various fits and starts and exploratory lawsuits, the FCC finally issued a comprehensive Internet regulation. It was called the Open Internet Order, or OIO, and it consisted of just three rules.²³

The OIO was the fruit of over half a decade of labor at the FCC, drawing not only on Powell’s principles, but also on over 10,000 comments from the public. The OIO and its later companion guide²⁴ distilled all this into two fundamental, tightly-linked, axiomatic claims. For just as all of Euclidean geometry could be derived from his axioms, so the FCC believed that the three OIO rules could be justified and derived from these two claims. I call them the Openness Claim and the Virtuous Circle Claim.

1. The Openness Claim #

In the eight page guide to the Open Internet Order the FCC wrote:

“[T]he Internet has thrived because of its freedom and openness—the absence of any gatekeeper blocking lawful uses of the network or picking winners and losers online. Consumers and innovators do not have to seek permission before they use the Internet to launch new technologies, start businesses, connect with friends, or share their views. The Internet is a level playing field. Consumers can make their own choices about what applications and services to use and are free to decide what content they want to access, create, or share with others.”²⁵

This is actually a bold claim. It’s not obvious that it is true; after all, the Internet started out as a highly restricted Department of Defense project. It also assumes that this “thriving” is a good thing; there are many authoritarian governments and religious leaders who would disagree.

But it is also an extraordinarily useful claim for making public policy and laws. Economic, national security, constitutional, and many other important controversies turn out to be about whether, how, and how much to limit Internet openness. The Openness Claim provides a guiding principle with which to answer those questions: the kind and quantity of this openness must always be such that the Internet can continue to thrive, in the ways described.

This “openness” is some version of net neutrality, although it isn’t exactly clear which version. In other words, the Openness Claim is nothing more than the claim that net neutrality is necessary for the success of the Internet.

2. The Virtuous Circle Claim #

The Openness Claim states that openness is necessary for the success of the Internet; but it does not say why. That is also explained in the companion guide to the OIO, in what I’ll call the Virtuous Circle Claim:

“This openness promotes competition. It also enables a self-reinforcing cycle of investment and innovation in which new uses of the network lead to increased adoption of broadband, which drives investment and improvements in the network itself, which in turn lead to further innovative uses of the network and further investment in content, applications, services, and devices.”²⁶

In Verizon v. FCC,²⁷ the FCC described this “self-reinforcing cycle” as a “virtuous circle”; hence my name for this second claim. It is an uncontroversial, mainstream view.

Finally: a mission! #

With these two claims, the FCC could finally justify its existence with respect to the Internet. Its Internet mission in life was now to preserve Internet openness, in order to preserve the virtuous circle. And (surprise, surprise) the civil servants of the FCC, if not always all of their five commissioners, did indeed believe there were threats to both Internet openness and the virtuous circle. To combat them, the Open Internet Order laid down three deceptively simple rules as law.

1. The Transparency Rule #

Internet service providers, or ISPs, had to be transparent. In order for competition to continue to drive up the speed and quality of users’ Internet connections, ISPs had to provide statistics and other data to consumers that would allow them to shop and compare. For example, customers had to be able to tell whether their Internet service was unreliable because their ISP was, or for reasons beyond their ISP’s control.

2. The Blocking Rule #

ISPs couldn’t block any lawful website or Internet connection.

3. The No Discrimination Rule #

ISPs couldn’t discriminate between websites or any other Internet connections. For example, if Comcast sold the same streaming movies as the non-ISP Netflix, it would be illegal for Comcast to slow down Netflix for all Comcast customers. Conversely, Netflix couldn’t pay Comcast to make Netflix movies stream faster.²⁸ What’s more, ISPs couldn’t start selling access to only part of the Internet; bundling was forbidden.

Exceptions to these rules were only allowed that were “reasonable” and not for profit, such as to protect a network from attack by viruses, for general network maintenance and optimization, as ordered by law enforcement, and so on.

The second and the third OIO rules were specifically designed to enforce the principle of net neutrality. The non-discrimination rule just makes Tim Wu’s version of net neutrality the law: an ISP can discriminate between packets within its networks, but only for reasons that are “reasonable”. The blocking rule is close to a special case of the non-discrimination rule. If you can’t discriminate between packets, you also can’t block some packets but not others. That is almost what the non-blocking rule says, but it goes further: an ISP cannot block any packets, period (except as required by law enforcement, or to protect its network, etc.).

In short, net neutrality was now the law of the land.

Verizon v. FCC #

On January 20th, 2011, less than a month after the Open Internet Order was published, Verizon sued to overturn it. As long as that suit was in court, the OIO was in a kind of legal limbo. Nobody knew whether it was law or not, and couldn’t until the judges issued their verdict.

The court case was called Verizon v. FCC, and it was decided²⁹ after three years by the United States Court of Appeals for the District of Columbia, on January 14th, 2014. It was this decision that ultimately led to the public uproar over net neutrality that year.

Even though Verizon “won” the case, the court’s three judges agreed with the FCC on virtually everything. They let the first OIO rule, the Transparency Rule, stand. They agreed that net neutrality was critical to the success of the Internet. They agreed that it was under threat. They agreed that the OIO was good policy: they agreed that it was a good way to protect net neutrality. Specifically, they agreed that it was good for net neutrality and the Internet to require that ISPs could not “reasonably” discriminate between or block packets. And they rejected all of Verizon’s challenges to the legality of the OIO — except, reluctantly, for one.³⁰

The one fly in the ointment that the court found was that the Open Internet Order could not apply to Internet service providers — which of course were really the only kind of company the FCC wanted to regulate in the first place. The problem, the court said, was that as written, the Open Internet Order could only apply to companies the FCC has classified as common carriers. The FCC classifies Internet service providers instead as information service providers, so the court said the Open Internet Order didn’t apply to them.

Which raises the question: what is a common carrier?

Common Carriers #

A key demand of vulgar net neutrality proponents is that the FCC reclassify ISPs as utilities, like today’s water, electricity, and telephone companies. The Internet, they argue, has become as essential and important to modern life as any of those public goods. Indeed, it is hard to read the FCC’s own eloquent description of the Internet without feeling the same way:

“The Internet is America’s most important platform for economic growth, innovation, competition, free expression, and broadband investment and deployment. As a “general purpose technology,” the Internet has been, and remains to date, the preeminent 21st century engine for innovation and the economic and social benefits that follow.”³¹

The legal issue in Verizon v. FCC, however, is not whether ISPs should be utilities; it is whether they should be common carriers.

If you don’t know what a common carrier is, you can relax now: you never will. That’s because it has never been defined properly.

Title II of the Communications Act of 1934 defines “common carrier” circularly:

COMMON CARRIER.—The term “common carrier” or “carrier” means any person engaged as a common carrier for hire, in interstate or foreign communication by wire or radio …

No further definition or description is ever given. Meanwhile, Title 47 of the Code of Federal Regulations of the United States of America unhelpfully defines a common carrier as whoever (not whatever) the FCC says is a common carrier.³²

The reason for all this vagueness appears to be that “common carrier” and “common carriage” are well-known legal terms, so legal documents don’t try to define them: anyone who is a lawyer is already supposed to know what they mean.³³ Unfortunately, there’s no authoritative legal definition anywhere either. For example, in 1901, the Supreme Court declared that a telegraph company is a common carrier under common law. This means that no statute needs to define the term; instead, the Court referred everyone to centuries of amorphous “common law” decisions.

What emerges from those centuries is an indistinct category of “general public” services, and a slowly growing number of obligations on them. You couldn’t discriminate; if you were a 17th century innkeeper, you had to offer a vacant room to anyone. Your goods were held to a higher standard than others; a common carrier and its products were often licensed and regulated, for example bakers and bread. To ensure general public access, you might well be required to provide services to your competition; telegraph and telephone companies had to share their lines, although they could charge each other for them. A common carrier might be a utility, like a telephone company, or it might not, like a public bus company.³⁴

In short, philosophers and lawyers cannot argue over whether Internet service providers are or are not a common carrier: no definition to argue over exists, anywhere. Instead, it’s up to the FCC to decide whether to classify ISPs as common carriers; and here the only thing the FCC commissioners, lawyers, and lobbyists argue about it is whether this would be good or bad for the Internet.

And they’ve been arguing about this for decades. Since at least 1998, the FCC has repeatedly declined to classify ISPs as common carriers, which include, for example, landline telephone companies. Common carriers must obey costly and onerous regulations listed under Title II of the Communications Act of 1934. ISPs have always been classified as “information providers”, which are subject to the much weaker regulations listed under Section 706, Title I of the Telecommunications Act of 1996. When you hear people arguing over “reclassification”, whether to “reclassify ISPs under Title II”, or “Title II vs Section 706”, or “Title II vs Title I”, this is what they are talking about.

The January 2014 Verizon v. FCC decision backed the FCC into a very tight corner. Either the FCC could give up on the OIO, or it could reclassify ISPs as common carriers, unleashing a political and legal firestorm from Republicans and industry.

For five months after the court’s 2014 decision in Verizon v. FCC, everyone waited to see what the FCC would do. And then, on May 15, 2014, the FCC proposed a way to keep the Open Internet Order without reclassifying ISPs. It essentially proposed to add a single word to the OIO:

“Commercially”.

“Commercially” #

In its Verizon v. FCC decision, the court pointed to a particular word used in the text of the Open Internet Order’s (OIO) net neutrality rules: “reasonable”. Companies providing Internet access, ISPs, had to be packet neutral: they could not treat different packets differently, except where this was “reasonable” — for example, to stop spam, to protect its network from attack, as required by law enforcement, and so on. The real point is that that list of “reasonable” exceptions does not include fees: an ISP cannot give preferential treatment to some packets in return for a fee.

In legal terms, “reasonableness” is a legal standard, like the “beyond a reasonable doubt” legal standard used in murder trials. The court said this “reasonableness” standard could only apply to “common carriers”, not to “information service providers”. The FCC, however, classified Internet service providers as the latter; so, the court said, the OIO did not apply to them, and therefore Verizon was off the hook.

The legal distinction between “reasonable” and “commercially reasonable” comes from a 2012 case, Cellco Partnership v FCC, decided by the same court, that asked whether cell phone companies have to provide roaming not just for voice calls (which was already legally required), but for data like the Internet and SMS. The FCC had issued a data roaming rule requiring cell phone companies to provide roaming under certain conditions, and Cellco sued, saying a cell phone comapany should not have to. The court had three options:

1. Cell phone companies don’t have to provide data roaming. For example, if you used AT&T, and you ended up in an area served only by Verizon, Verizon wouldn’t have to let you use its data network if it didn’t want to.

2. Cell phone companies must provide data roaming, at no extra charge above voice roaming; the only exceptions would have to be “reasonable”, as in to prevent spam and so on.

3. Cell phone companies must provide data roaming, but they can charge for it: prices just have to be “commercially” reasonable. In other words, the cell phone companies couldn’t try to weasel out of offering data roaming by making it so expensive nobody would want to pay for it.

The court chose the third option because, they said, the FCC data roaming rule had 16 good criteria and a “catchall”, which together amounted to a more flexible “commercially reasonable” standard. That is why roaming charges in America are now quite high.

So in Verizon v. FCC, the court wrote that if the OIO rules were changed to allow “commercially reasonable” packet discrimination, as in letting Netflix pay an ISP to deliver its movies faster, then the rules “might” no longer be common carrier rules.³⁵

“Might.” Maybe. In other words, if the OIO were changed that way and Verizon sued again, the court might side with the FCC. No guarantees.

In the end, the FCC never made up its collective mind about what to do.

On May 25, 2014, the FCC published a “Notice of Proposed Rulemaking”³⁶ that proposed to add the word “commercially”; but it had been forced into publication over the protest of most of the FCC’s four commissioners by the Democrat chairman Tom Wheeler.

Commissioners Mignon L. Clyburn and Jessica Rosenworcel had no discernible opinion. In the Notice of Proposed Rulemaking they wrote stirring declarations of their commitments to justice and openness, awed praise for the Internet and America, assurances that they were listening to the American people, and then disagreed with the rulemaking proposal without saying why. Rosenworcel said she disagreed with it, even though she concurred with it; Clyburn didn’t say where she stood at all, contenting herself with a mere “statement”. Commissioner Ajit Pai’s opinion was that the FCC should let Congress decide what to do.³⁷

Net neutrality was a political problem for each of these three commissioners, nothing more. As with gun control or abortion, it was impossible to stake out a position without being mercilessly attacked; but in the case of net neutrality, it was not yet clear which position was the safest political ground. So they presented themselves as actively engaged and deeply concerned with the issues, and left it at that.

The FCC’s Options #

When the Verizon v. FCC decision was handed down, defeating the FCC’s attempt at making a net neutrality law, the FCC was left with essentially four options.

1. Do nothing: accept defeat, accept that net neutrality was not going to be law, and then continue to putter along protecting the Internet on an ad hoc, case-by-case basis. The Telecommunications Act still authorized the FCC to do that; it would just be more difficult without the Open Internet Order.³⁸

2. Add “commercially” to the OIO: Tweak and weaken the Open Internet Order regulations so that they would apply to “information service providers” like ISPs after all, and then hope that if this new OIO was challenged again in court, it would be affirmed.

3. Go with Title II: Reclassify ISPs as “common carriers”³⁹. The court had said this would make the OIO apply to ISPs after all. The risk was that when the ISP industry inevitably sued to overturn this reclassification, they would win.

4. The “Third Way” (what I call “Title II ‘lite’): reclassify ISPs as "common carriers”, but only in name. The FCC would promise not to enforce the many onerous rules other common carriers are subject to. This would hopefully make it less likely that ISPs would sue to overturn the reclassification, or if they did, that they would win.

We are going to take a close look at each of the four options the FCC faced, and ask whether they would foster the virtuous circle, undermine it, or have no appreciable effect on it. Recall that net neutrality is only a means to an end, and that end is the virtuous circle: users “vote” for Internet products by using them so much, they strain the Internet’s network capacity, which leads to Internet service providers and others improving their networks, which in turn leads to the possibility of even more network intensive products that users may “vote” for, and so on ad infinitum.

Vulgar net neutrality has led many people to believe the choice is obvious. This is because vulgar net neutrality ignores the underlying engineering. We won’t, and as a consequence, we will see why the choice is actually hard.

FCC Option #1: The Do Nothing Option #

After a federal court invalidated the FCC’s net neutrality law, the Open Internet Order (OIO), in January, 2014, the FCC had the option to do nothing: to give up on making net neutrality the law, and revert back to the status quo ante. One way to decide whether this would have been reasonable is to look at the FCC’s reasons for passing the OIO in the first place.

The number of actual verified violations the FCC listed in 2010, and again in 2014, is ludicrously small, too small to justify - as the only reason - a sweeping net neutrality law. Of course, there may be many more violations; the FCC lists about two dozen reports of violations, some of them in foreign countries, that it was not able to investigate itself; the FCC has limited investigatory powers.

As the FCC described them, these few cases do not sound like net neutrality violations at all, but like Internet service providers trying fairly reasonably to protect their overloaded networks. Look beyond the FCC’s own reports, however, and we discover that it was right: these were serious net neutrality violations, sometimes for reasons of subtle engineering.

Most importantly, though, the bulk of the FCC’s evidence is about ISP’s incentives and ability to violate net neutrality, not actual violations. This should be especially troubling to us as Americans, in light of very recent history. Let us remind ourselves why.

Weapons of Mass Neutrality #

In the contentious year leading up to the American invasion of Iraq in 2003, everyone could agree at least on this: that Saddam Hussein had compelling and powerful incentives to build and stockpile weapons of mass destruction. He wanted to hold off Iran with them, a country with whom he had already fought, and lost, the longest conventional war of the twentieth century; to kill his own people en masse with them, something he had done once before, in Halabja; to intimidate and coerce the many nations Iraq shares borders with; to fulfill his own sense of grandeur, beyond the many palaces and monumental sculptures he had already built. The list goes on, and it is long.

Everyone also agreed that Saddam had the capability to build nuclear weapons; after all, just twenty years earlier, Israel thought he got so close that it destroyed his Osirak nuclear reactor.

People did disagree over whether Saddam actually had weapons of mass destruction; and within a year of the invasion, it became clear that he didn’t. Not one. Not a sausage. It wasn’t for want of trying, however, although most of his attempts dated from before 1991. Of the few efforts made before 2003, all were successfully prevented by sanctions and inspections, without the need for an invasion or a war.⁴⁰

There is a sobering parallel with the FCC’s arguments for why the Open Internet Order is necessary: they too depend almost exclusively on arguing that ISPs have the incentive and the capability to violate net neutrality, rather than on hard examples of such violations.

Absence of evidence, evidence of absence #

In its May, 2014 “Notice of Proposed Rulemaking”, the FCC writes proudly, and at some length, about how in Verizon v. FCC the court agreed with the FCC about incentives IPSs have to act non-net neutrally (emphasis added):⁴¹

The D.C. Circuit found that the Commission’s assessment of broadband providers’ incentives and economic ability to threaten Internet openness was not just supported by the record but also grounded in “common sense and economic reality.” It affirmed the Commission’s conclusions that vertically integrated broadband providers⁴² have incentives to interfere with competitive services and that broadband providers generally have incentives to accept fees from edge providers.⁴³ And the court cited with approval the Commission’s conclusion that a broadband provider⁴⁴ would be unlikely to fully account for the harms resulting from such practices. The court also upheld the agency’s conclusion that such incentives could “produce widespread interference with the Internet’s openness in the absence of Commission action.”⁴⁵

So, the court agreed with the FCC that ISPs have incentives to violate net neutrality. What the FCC did not report the court as saying, because the court never said it, was that companies actually were violating net neutrality in significant numbers. In fact, those numbers were not significant. The ones that the FCC itself could verify amounted to a single digit.

The number of verified net neutrality violations #

On December 23rd, 2010, the FCC announced its net neutrality regulation, the Open Internet Order, in a document called a “Report and Order”. Of the 174 numbered paragraphs it contained, only two, paragraphs 35 and 36, were concerned with whether any actual violations of net neutrality had already taken place.

Between 1934 and 2010, the number of those incidents was exactly three; a fourth was only “alleged”. The number of examples of violations that only might have violated net neutrality (“We do not determine in this Order whether any of these practices violated open Internet principles …”) was exactly five.

By 2014, three to four years later, the number of additional violations the FCC had verified was exactly two; it reported them in its May, 2014 “Notice of Proposed Rulemaking”.

Thus the total number of net neutrality violations the FCC has ever verified and reported comes to 5, scarcely enough to justify a new law to enforce it. The nature of those violations, however, tells a more troubling story.

The nature of verified net neutrality violations #

Of the five incidents, one was probably unrelated to net neutrality,⁴⁶ and another was a rather transparent and hairbrained violation by a relatively small ISP;⁴⁷ the other three, however, were serious.

The Comcast/Bittorrent net neutrality violation (2008) #

In 2008, the FCC ordered Comcast to stop blocking Bittorrent software used by some of its subscribers. Bittorrent is used to download large files quickly, and while it is often used for pirated movies and software, there are many legitimate uses as well, such as downloading Linux operating systems and other open source software.

The FCC’s brief descriptions of the incident, in the Open Internet Order and again in the May, 2014 Notice of Proposed Rulemaking, give the impression that Comcast was merely trying to stop a handful of users clogging the entire Comcast network with illegal downloads. But investigations by the FCC and the Electronic Frontier Foundation had proved otherwise. Comcast was doing this secretly, and to all its Bittorrent users regardless of how little they downloaded. Comcast didn’t try to single out only the worst abusers. Comcast’s actions also affected many users outside its network, users who didn’t subscribe to Comcast, many of whom abandoned Bittorrent in the mistaken belief that Bittorrent didn’t work. Comcast was breaking the virtuous circle; we may never know what new breakthrough startups might have been built around Bittorrent if this hadn’t happened.

The FCC ordered Comcast to stop, but Comcast sued, and in 2010 it won. A direct outcome of this setback was the FCC’s introduction of the Open Internet Order at the end of that year; it’s Non-Blocking Rule was targeted directly at Comcast.

The AT&T Facetime and Verizon tethering violations (2012) #

In 2012, Apple introduced a high-definition video telephone product called Facetime for the iPhone. Almost immediately, AT&T announced it could only be used on its more expensive cell phone plans. This was a blatant case of bundling, of only offering an Internet product at a premium, and the FCC quickly put a stop to it.⁴⁸ That same year, the FCC fined Verizon $12.5 million for blocking all third-party tethering apps, which would have allowed users of Android and Apple phones to use their phones as a wifi hotspot. Tethering was allowed only for more expensive cell phone plans, using a Verizon product. This too was a clear violation of net neutrality, for the same reason, but it took the FCC the better part of a year to browbeat Verizon into submission, without taking them to court.⁴⁹

Net neutrality violations in Europe #

In part because it could find so few net neutrality violations in the United States, the FCC also tried to argue that net neutrality was being violated in many other countries, for example in Europe:

… a survey conducted by the Body of European Regulators for Electronic Communications (BEREC) shows that European Internet service providers reported engaging in … blocking and throttling …⁵⁰

That 2012 report is available online here.⁵¹ In figure 3, shown below, the sixth and seventh bars from the left show how many of 32 European countries’ ISPs engaged in blocking and throttling: less than 4%, or 10 out a total of 266. The remaining bars in the figure tell the same story: in Europe, where there are currently no net neutrality regulations, violations of net neutrality are negligible.

What was the right choice? #

Reasonable people can disagree over whether the FCC’s evidence justified the Open Internet Order, and whether it ought to abandon it.

On the one hand, it only verified five cases of ISPs violating net neutrality, not all of them convincing; and although it listed dozens more that it did not investigate, in one case at least, European violations, it either failed to actually read the source cited, or deliberately misrepresented the source as reporting a significant number of offenses. On the other hand, three of its verified cases were serious.

The FCC can’t stop Comcast from blocking Bittorrent; it lost that case in court, and introduced the Open Internet Order in response. Comcast appears to have stopped all Bittorrent blocking because of the bad press it received, but currently there is no law to stop it from resuming.

As for the other two cases involving Verizon and AT&T, the FCC managed to settle them without going to court, but we simply do not know whether this was perhaps because the Open Internet Order case, Verizon v. FCC, had not yet been decided. The two cases occurred in 2012; the Open Internet Order was only reversed in the courts in 2014.

Given all this, if the FCC decides to abandon the Open Internet Order and any other dedicated net neutrality regulation, should we expect the ISP industry to begin to seriously damage the virtuous circle? Would the FCC be able to keep them in check with the Telecommunications Act alone? In short, should the FCC’s response to the Verizon v. FCC decision have been to do nothing?

It’s a difficult call. Personally, I’m inclined to say “no” because the three cases seem quite ominous to me; but that’s just me, and frankly, I feel a bit silly that we’re only talking about three cases (but don’t forget, the FCC didn’t have the resources to verify the many other cases it reported second-hand).

The only one of the five FCC commissioners to seriously address this issue in the May, 2014 “Notice of Proposed Rulemaking”, other than the chairman Tom Wheeler, was Commissioner Tim O'Rielly;⁵² and his answer was “yes”, for many of the same reasons. He may well be right.

FCC Option #2: Add “commercially” to the Open Internet Order #

After a federal court struck down the Open Internet Order (OIO) in January, 2014, the second of the FCC’s four options a was to revive it by essentially adding the word “commercially”; for brevity, I’m going to refer to this option as OIO+c.

The original Open Internet Order did not allow ISPs to block or discriminate between Internet packets, except on “reasonable” grounds such as stopping a computer virus, obeying law enforcement, and so on. Profit was not reasonable grounds. The court ruled that this legal “reasonableness” standard meant the Open Internet Order could not apply to Internet Service Providers; but the court suggested that if profit was allowed, if the Open Internet Order were revised so that exceptions could be “commercially reasonable”, then they might let it stand.

In May, 2014, the FCC proposed to replace the Open Internet Order with OIO+c, and gave the public three months to comment.

The original Open Internet Order was designed to protect the virtuous circle of the Internet; the question was, would OIO+c protect the virtuous circle, destroy it, or have no appreciable effect on it?

A majority of the public believed that OIO+c would destroy the virtuous circle, because it would legalize Internet “fast lanes”. The number of public comments broke all FCC records, and at times crashed its website; most were angry, even abusive.

When viewed through the lens of vulgar net neutrality, this seemed outrageously obvious. But vulgar net neutrality is wrong. It ignores how the Internet is actually engineered, and it ignores what the FCC actually said. Unfortunately, once we resolve those problems, we are left with many possible judgments about whether or not OIO+c would undermine the virtuous circle; and as we will see, none of them is wholly convincing.

Commercially unreasonable #

OIO+c would allow Internet service providers to charge companies like Netflix a “commercially reasonable” fee to discriminate in favor of their packets, giving them preferred treatment over packets from other companies.

But what does this mean? What would make such a fee commercially unreasonable? It is clear what that means when it comes to data roaming fees: cell phone companies can charge for roaming, as long as the price is not so high that most people cannot afford it. But what does “commercially unreasonable” mean under OIO+c?

The court didn’t say in its Verizon v. FCC decision, and the FCC isn’t sure either. The FCC noted that the court characterized “commercially unreasonable” as “[c]onduct that unreasonably restrains trade …”, but the FCC couldn’t give a single definite example of what that might be. Instead, it almost plaintively asked the public for help: “Are there any practices that, consistent with the Verizon court’s reasoning, could be viewed as per se commercially unreasonable?”⁵³

In the meantime, the FCC has proposed to interpret “commercially unreasonable” as fees and financial practices that would damage the virtuous circle:

“… we tentatively conclude that the Commission should adopt a rule requiring broadband providers to use “commercially reasonable” practices in the provision of broadband Internet access service … [The rule] would prohibit as commercially unreasonable those broadband providers’ practices that, based on the totality of the circumstances, threaten to harm Internet openness and all that it protects.”⁵⁴

That’s about as far as the FCC got; beyond that, they say, “… we seek comment on how we should construct factors in applying the commercially reasonable legal standard…”.⁵⁵

And people have commented. Two types of putatively commercially unreasonable conduct have emerged; one involves “fast lanes”, the other type does not. Let’s look at this last kind first.

“Commercially unreasonable” practice #1: subsidized unlimited streaming #

Under virtually every cell phone data plan, there is a monthly data allowance, or “cap”, on how much data you can download. If you go over your monthly cap, then depending on your provider, your download speed will be slowed, or you will be billed extra “overage” charges, or you will be cut off until the next billing cycle.

For example, a typical cell phone cap is 2 gigabytes per month. If 1 gigabyte is the size of a 90 minute Netflix movie,⁵⁶ we could watch two movies a month on our phone before we “hit” our cap. If we watch two more movies, or download 2 gigabytes in some other way, our cell phone bill might double.

Home Internet downloads have monthly caps as well, but these are usually from about 250 to 600 gigabytes.

Beats is a music streaming service that costs $15/month. In early 2014, AT&T introduced unlimited streaming only for Beats: AT&T subscribers can stream an unlimited amount of music from Beats every month. If you are an AT&T user, this is obviously a strong incentive to use Beats over its competitors Pandora, iTunes Radio, Spotify, and others; Beats paid AT&T a fee to make this happen.⁵⁷ T-Mobile responded with a similar, broader program called “Music Freedom”: any music T-Mobile cell phone users stream from a handful of Beats’ competitors will not be counted towards their monthly data cap. Those services all paid T-Mobile a fee to be part of Music Freedom.

Any startup trying to get into the music streaming business is now at a disadvantage. Even if it offers better music, better software, and better streaming quality than Beats and Pandora and the rest, AT&T and T-Mobile users still won’t want to use it: too many will prefer unlimited music to a better music experience. And a startup probably can’t afford the fees AT&T, T-Mobile, and others are charging for unlimited streaming. The virtuous circle has been broken: money, not quality, will now decide the online music winners.

This is a clear violation of both the spirit of net neutrality, and of all three versions of it that Tim Wu defined in 2003. Currently, however, no law exists that makes this kind of unlimited streaming illegal. The question is, would unlimited streaming be illegal under OIO+c?

The FCC noted the emergence of these programs,⁵⁸ and asked the public to comment on them. The FCC has proposed to interpret “commercially unreasonable” as anything that damages the virtuous circle, so presumably the FCC would find unlimited streaming illegal under OIO+c; but the courts might not agree. After all, OIO+c also explicitly permits fees for packet favoritism.

Alternatively, the FCC might try to prevent these program by claiming that they violate the Communications Act of 1934. This would be more difficult than arguing that they violate OIO+c, and again, the courts might not agree.

Nobody can say for sure. As a test case for or against adopting OIO+c, unlimited streaming is not clear-cut.

“Commercially unreasonable” practice #2: fast lanes #

The FCC has asked the public for examples of Internet service behavior which threaten the virtuous circle, and to comment on whether the OIO+c standard of “commercially unreasonable” could legally prevent them. Two kinds of examples have emerged. Whether OIO+c would prevent the first kind, which is to charge Internet companies like Beats a fee in return for letting users download unlimited amounts of its data, is simply unclear.

The second kind involves Internet service providers charging Internet companies a fee in return for giving users limited but faster downloads: charging a fee for “fast lanes”.

Fast lane type #1: caching fast lanes #

There are two kinds of fast lanes. I described my experience of the first “caching” kind earlier, in Indonesia. Web sites 10,000 miles away in the United States loaded excruciatingly slowly, except for ones like cnn.com that had paid a company to store, or “cache”, their data on computers much closer to Indonesia.

Caching can happen without an Internet service provider discriminating between packets, or even knowing that caching is going on. One famous caching company, in fact the first one ever, is Akamai. An Akamai salesman sitting in Cambridge, MA, can call up an Internet service provider in Indonesia and request an Internet connection for an Akamai “cache” computer sitting in the capital, Jakarta. The ISP will not know or care what the computer is for; all that matters is how much data will flow in and out of that computer, and that Akamai pays for that data on time. Once a contract is signed, the ISP will just want to know where in Jakarta to send its cable guy.

When I sat on my beach in Indonesia and pointed my web browser at cnn.com, it first sent a tiny handful of packets 10,000 miles away to Atlanta, GA, where cnn.com’s servers are located, via the Indonesian ISP. A server in Atlanta then sent a tiny handful of packets back 10,000 miles to my browser with the instruction, “go ask the following server in Jakarta instead”; and my browser did. That server, only a few hundred miles or less away, sent an enormous number of packets to me, with cnn.com’s home page, images, and video. The distance was hundreds of times shorter than 10,000 miles, so it didn’t take as long. During all these packet exchanges, the Indonesian ISP blindly conveyed packets back and forth; it treated the packets the same as any others. It didn’t violate packet neutrality.

Caching fast lanes haven’t compromised the virtuous circle, mainly because there are many caching companies to choose from. Market competition keeps prices down, making them affordable to everyone from small startups to large companies like cnn.com.

The original Open Internet Order explicitly allowed these kinds of fast lanes, and so would OIO+c.⁵⁹ Not only have they been around for over a decade without damaging the virtuous circle, they do not violate any of the kinds of net neutrality Tim Wu identified. Caching fast lanes are not a net neutrality issue.

Fast lane type #2: direct interconnect fast lanes #

On February 23, 2014, Netflix and Comcast announced a so-called “direct interconnect” agreement. Less than a month later, on March 20, 2014, the chairman of Netflix, Reed Hastings, unexpectedly attacked Comcast over the deal on his company’s blog; and yet, in the coming months, he closed similar deals with Verizon and AT&T. None of the parties has ever disclosed the terms of the deals, or how they came about.

In each case, Netflix paid the Internet service provider; what we don’t know is why or how much. There are many different accounts that all contradict each other, and that disagree about what kind of fast lane Netflix purchased, or whether it is even fast. The issue isn’t whether OIO+c would prevent these fast lanes; the issue is whether they even exist.

The John Oliver version #

The most famous version was popularized by the late-night comedian John Oliver in a skit on June 1st, 2014. In it, he displayed a graph he got from Netflix’s web site (see below). Netflix streaming deteriorates steadily during the last months of 2013, then suddenly and dramatically improves around the time of the deal. Oliver then says: “See if you can guess when Netflix agreed to Comcast’s demands.”

As John Oliver tells it, this was a “shakedown”. Comcast allowed Netflix streaming to deteriorate until Netflix agreed to sign a special deal with Comcast for a “fast lane”. Oliver then went on to explain that these fast lanes would destroy the virtuous circle, that the FCC was proposing to legalize them, and he encouraged everyone, especially Internet trolls, to submit comments to the FCC denouncing them.

How the Internet works #

Like all vulgar net neutrality stories, this is a good one, with clear villains and victims and plenty of outrage; and like most of them, it makes no engineering sense. To understand why, we first need to learn how networks on the Internet make money.

The Internet and nuclear war #

Note: If you already know about Internet packets, how they travel across the Internet, and about settlement-free and paid peering and transit, you can skip this section.

One of the design goals of the Internet was to be able to survive a nuclear war, even a world-wide one. When we send a message over the Internet, whether it is an email, our voice over Skype, a web page, or a movie from Netflix, the data doesn’t travel along a wire like a freight train. Instead, it explodes like a hand grenade into little fragments of data called packets. Each packet makes its separate way along the Internet. It is sent from computer to computer, like hopping from stone to stone to cross a stream. The packets are all numbered, so that the receiving computer can tell when it has received all the packets of a message, and in what order to reassemble them. There are many different paths between most two points on the Internet, just as there are many different ways to drive from one American city to another. Whenever there is a choice between paths, a guess is made as to which path will be fastest, and a packet is sent that way. The guesses vary with time, based on congestion and “traffic jams”. Like water flowing around obstacles, the packets always find a way, as long as the network is big enough and interconnected enough. If a nuclear strike takes out parts of the network, the packets are simply rerouted on surviving paths.

Tiers and backbones #

Take a few computers, connect them with wires so they can talk to each other, and we have created a computer network. Connect millions of such networks together, and we might end up with something like the Internet: an interconnected network of networks.

The actual Internet is more organized. At its core is a small collection of competing private companies called backbone providers. Each backbone provider owns a superhighway to nowhere, with on and off ramps, essentially a single cable that encircles the world. A backbone provider’s cable goes under oceans, through mountains, and across continents, but it doesn’t really connect to any websites. Instead, it connects networks, typically large networks like Comcast and Verizon.

When we subscribe to Comcast’s Internet service in Cambridge, MA, we become one more computer connected to Comcast’s very large computer network. Comcast guarantees that we will be able to connect to any website on the Internet, anywhere in the world. There are two ways it could fulfill that promise. Either it could run cables from the Comcast network to every single other Internet network on the planet, of which there are millions; or it could connect to a backbone provider. Comcast, like all its competitors, has chosen the latter.

A backbone provider doesn’t run cables to every last network on the Internet either. Instead, a tiny network connects to a larger network, which connects to a larger network, and so on up the food chain, until finally a large network connects to a backbone. It’s the free market at work, not a central planner. People talk about “tiers” of providers; a backbone is a tier 1 provider. And even though a backbone connects computer networks rather than individual computers, it is still also a computer network.

How Internet networks make money #

There are many different kinds of networks on the Internet, so it is always dangerous to attempt to make statements that are true of all of them; but broadly speaking, they make money through “peering” and “transit” agreements.

As the following figure shows, peering is when two networks connect and exchange traffic, but the traffic is always between computers inside each of their two networks. Sometimes that traffic is about equal each way most of the time, and the two networks don’t charge each other for it; that’s called settlement-free peering. Sometimes there’s more traffic going one way rather than the other, and one network will charge the other for it; that’s called paid peering. In some agreements the networks carefully measure the flow and settle up at the end of every month, in other cases they agree on a fixed monthly payment.

In a transit agreement, a network is paid for data that crosses it in transit to somewhere else. Transit is typically billed monthly, and always to the neighboring network the traffic came from.

These agreements between two networks can be a legal contract, but it is not uncommon for them to be an informal unwritten arrangement.

Peering and transit agreements are illustrated schematically in the figure below.

The Comcast/Netflix deal #

Until early 2014, when a Comcast subscriber sat at a computer at home and clicked a link to view a Netflix movie, that sent a few packets through Comcast’s network, across some backbone which Comcast paid transit to, and to Netflix’s computers. The movie then travelled across a backbone (not necessarily the same one as before), for which Netflix paid transit; into Comcast’s network, for which the backbone paid Comcast peering; and onto the subscriber’s computer.

In short, backbones acted as middlemen between Netflix and Comcast, earning transit. Comcast earned paid-peering from the backbones. Netflix earned money from its movie subscribers. All appeared well.

Then, on February 23, 2014, Netflix and Comcast announced that henceforth Netflix’s movie servers would connect directly to Comcast, instead of through a backbone. This was really just a paid peering agreement, nothing more, with Netflix paying Comcast; it was also called a “direct interconnect” agreement.

Within two months Netflix closed similar deals with Verizon and AT&T.

At the time, neither side revealed how much Netflix was paying, who had initiated the agreement, or why. Nobody knew the full story; but before long, there were many different versions. All but one of these versions claims (1) that the deal was for a fast lane, (2) that it violated net neutrality and endangered the virtuous circle, and (3) that OIO+c would encourage or legalize such deals.

When confronted with the reality of how the Internet is engineered, each of these versions crumbles into incoherence, or raises more questions than it answers. We will name and analyze the most prominent ones below, which I have numbered as “theories” one through six. The most plausible explanation of the Netflix/Comcast deal has no villains and has nothing to do with net neutrality. I call it the Bad Backbone version; we will look at it last.

Theory 1 of 6: The Paid Peering Blackmail version #

On March 20, 2014, less than a month after the Netflix/Comcast deal was, Netflix CEO Reed Hasting’s wrote a blog post arguing that Comcast shouldn’t charge anyone to connect to it. It is utterly unclear what exactly he was claiming. On the one hand, he never used the word “peer” or “peering” except in a footnote; instead, he complained about “interconnection fees”, a term nobody uses for peering. On the other hand, he seemed to be arguing that Comcast should provide settlement-free peering to everyone. He called this principle “strong net neutrality”:

Strong net neutrality additionally prevents ISPs from charging a toll for interconnection to services like Netflix, YouTube, or Skype, or intermediaries such as Cogent, Akamai or Level 3,⁶⁰ to deliver the services and data requested by ISP residential subscribers. Instead, they must provide sufficient access to their network without charge.

Hasting’s argued that Comcast made a profit from promising its home subscribers specific connection speeds, and since Comcast didn’t share any of that profit with Netflix, Netflix shouldn’t have to share the cost of upgrading Comcast’s network to ensure those speeds — even if Netflix traffic was slowing down the whole Comcast network. He then accused Comcast of blackmail. He said Comcast had refused to upgrade its Internet connections as Netflix’s popularity surged, causing increasingly poor Netflix service for Comcast home subscribers, until Netflix agreed to pay Comcast “interconnection fees” or “tolls”. He also argued that small startups couldn’t afford these “fees”.⁶¹ This was essentially the same version John Oliver presented over two months later.

Comcast responded that its agreement with Netflix was just like any of its other paid peering agreements with backbone providers; that Netflix had approached Comcast about cutting out the backbone middlemen; that Netflix had used its market power to extract a low price from Comcast for the agreement; and that anyway, if Netflix didn’t like their Comcast agreement, they were free to go back to using backbones instead, so why were they suddenly complaining?

Everything here turns on what Reed Hasting’s meant by “interconnection fees”. If he meant paid peering, his blackmail allegation makes no sense at all.

If Hastings is talking about paid peering, Comcast gets paid either way, whether Netflix uses a backbone, or directly connects to Comcast. Either way, Comcast gets paid via a paid peering agreement. So what motive would Comcast have to prefer a direct deal, let alone blackmail Netflix into one? Nobody, not even Hastings, is saying that Netflix is paying Comcast more than a middleman backbone.⁶² The only benefit to Comcast of a direct connection seems to be that they get to deal directly with their customer, instead of always going through a middleman — something Netflix must be happy about, too. But then why would Comcast blackmail Netflix to do something that Netflix would prefer to do anyway?

And if Hastings is talking about paid peering, his other arguments also make no sense. Paid peering has been around for decades, practically since the dawn of the Internet, and it didn’t stop Netflix, Facebook, Google, and Skype transitioning from small startup to behemoth: so why should it stop startups now? The other argument, that since Comcast doesn’t share any of its profits with Netflix, and therefore Netflix shouldn’t have to share the costs of network upgrades, is bizarre. It implies that Comcast is already making money because of Netflix, because Netflix’s popularity attacts new Comcast subscribers and makes them willing to pay high Comcast fees. But what if Netflix becomes unpopular next month? Should Comcast then be allowed to charge Netflix paid peering after all? Should only big, popular Internet companies get settlement-free peering, and the unpopular ones have to pay paid peering? Whether Hastings likes it or not, his argument commits him to this absurd position.

In short, the Paid Peering Blackmailing/John Oliver version isn’t convincing at all; it is bewildering. We won’t ask whether OIO+c would prevent it, because it doesn’t make sense.

But what if Reed Hastings didn’t mean paid peering?

Theory 2 of 6: The Add-On Fee Blackmail version #

Perhaps Reed Hastings was confused. Perhaps what he meant to say was that Comcast wasn’t only charging for paid peering, Comcast was demanding an additional fee as well: an “access fee” or “toll”, just for the privilege of connecting to Comcast at all. Paid peering has been around for decades, but an additional access fee is unheard of. This is what a General Counsel at Level 3, a large backbone provider, had accused various unnamed ISPs of two days before Reed Hasting’s outburst, on March 18, 2014; three days later this General Counsel filed a complaint about these fees with the FCC. He said that even when backbone networks like Level 3 were willing to pay these Internet service providers (ISPs) to upgrade congested paid peering connections, the ISPs refused unless this additional “fee” was paid. He accused the ISPs of blackmail.

Unfortunately, the General Counsel refused to name names or ISPs, even to the FCC, and he didn’t say why. He just referred to “ISP X”, and his damning evidence was a single chart. If anyone else is independently complaining about an extra fee, I haven’t been able to find them. It’s true that at the time of the Open Internet Order, the FCC warned that something like this might happen; but it hadn’t yet.

The Add-On Fee Blackmail version is initially appealing in part because it would make Reed Hasting’s blog post a bit more reasonable. A connection fee on top of paid peering is unprecedented, and seems unjustifiable.⁶³ But we can’t be sure that this is what Hastings meant, and the evidence that Comcast or anyone else demands such “access fees” is suggestive, but insubstantial. Even if such fees exist, we don’t know whether they are large enough to harm the virtuous circle; for all we know, they may be commercially reasonable.

The real point, though, is that this version of the story fails to explain why Comcast would prefer a direct interconnect with Netflix, let alone be willing to blackmail Netflix for it. Comcast will get this alleged fee whether it is paid by a backbone intermediary, or by Netflix: why should Comcast prefer a direct interconnect, which Netflix prefers as well?

So maybe this add-on fee exists, and maybe it doesn’t. Even if it exists, it doesn’t explain the direct interconnect deal.

Would OIO+c make such a fee illegal? It’s difficult to say; it really depends on whether those fees undermine the virtuous circle — but nobody seems to know if the fees do, or if they even exist.

Theory 3 of 6: The Slow Lane Blackmail version #

Verizon is a cable television company as well as an Internet service provider. It wants its subscribers to buy more cable TV channels, and to buy movie rentals from the video on-demand service it owns. If these customers choose to watch Netflix instead, Verizon loses money. So Verizon has a perverse incentive to make Netflix perform poorly inside its network, to discourage Verizon subscribers from using it. The same can be said of Comcast, or of any ISP that also sells cable TV services.

The Slow Lane Blackmail version of the Netflix/Comcast deal has two parts: the Slow Lane part, and the Blackmail part. The Slow Lane part says that ISPs aren’t trying to sell or create fast lanes, they are creating slow lanes to harm competitors like Netflix: they are deliberately slowing Netflix traffic within their networks. The Blackmail part says that ISPs blackmailed Netflix into paying stiff fees and direct interconnect agreements in exchange for “unslowing” Netflix, and in the hope that Netflix will be forced to pass on those fees to Netflix customers by raising prices.

Before we look at the evidence, note that the Blackmail part doesn’t make a lot of sense. If ISPs really are successfully slowing Netflix traffic, there is no (obvious) reason for them to blackmail Netflix into doing anything: they’ve already accomplished their alleged goal. There’s no need for a complicated conspiracy theory of fees, direct interconnect, coordinating their efforts, and the rest.

Now let’s look at the evidence, first for the Slow Lane part, then for the Blackmail part; and then we’ll look at evidence against both.

There are two kinds of evidence that ISPs are deliberately slowing Netflix traffic. The first kind comes from Level 3, a backbone provider.

Theory 3.1: Level 3’s evidence of a Slow Lane #

In July, 2014, one of Level 3’s vice presidents, Mark Taylor, published a detailed account of how Verizon was refusing to upgrade a Level 3/Verizon paid peering connection. It was being swamped with ever increasing Netflix traffic, and as a result it was painful to even try to watch Netflix on Verizon’s network. The upgrade would have taken five minutes, cost almost nothing, and Level 3 had repeatedly offered to pay for all of it; Verizon simply refused. Taylor couldn’t prove why, he could only speculate:

Could it be that Verizon does not want its customers to actually use the higher-speed services it sells to them? Could it be that Verizon wants to extract a pound of flesh from its competitors, using the monopoly it has over the only connection to its end-users to raise its competitors’ costs?

That’s the Slow Lane Blackmail theory in a nutshell, except that Mark Taylor never says that Verizon actually demanded a “pound of flesh” fee, from Level 3 or anyone else. As we saw earlier (see the Add-On Fee version above), Level 3’s General Counsel had claimed this was happening, but he didn’t name names or numbers. Both men write angrily and persuasively, but ultimately they are just asking us to take their word for it: why? Maybe Level 3 is the bad guy in this story: maybe their paid peering agreement with Verizon is for less data than they are trying to push onto Verizon’s network. We just don’t know, and we don’t know because neither Level 3 nor Verizon will come clean with what they’ve got on each other. The evidence is suggestive that Verizon is trying to slow traffic through neglect, but it is hardly compelling. I’m not aware of any similar public charges against Comcast or AT&T.

Theory 3.2: The OpenConnect evidence of a Slow Lane #

For the past few years, Netflix has been offering boxes to all Internet service providers that will speed up Netflix for the ISPs’ subscribers — for free. Netflix calls this free service OpenConnect. Many smaller ISPs have installed OpenConnect boxes, but not the giants like Verizon or Comcast. Why not? Since OpenConnect is free, the only plausible reason is that it would compete with cable TV services Verizon and Comcast offer. OpenConnect wouldn’t be free at all for them, because people could watch shows and movies on Netflix instead of purchasing them from video on-demand, or buying more cable channels.⁶⁴

This is irrefutable evidence of inaction: Verizon and Comcast are not doing something that would improve, and possibly fix, Netflix performance inside their networks. But it’s also a little hard to blame them for not installing something which would only help their competition.

Theory 3.3: The evidence of blackmail (with a slow lane) #

There is no direct evidence that Comcast, Verizon, or AT&T blackmailed Netflix into direct interconnect agreements or anything else by slowing Netflix traffic. Reed Hastings, the voluble CEO of Netflix, has never accused them of this; he only accuses them of charging fees, which he never specifies, and which may be common-or-garden paid peering charges. Like John Oliver, everyone who makes these charges of blackmail simply assumes that Netflix didn’t want a direct interconnect agreement and so must have been blackmailed. And as I already pointed out, this inference doesn’t even make sense: why would these ISPs undertake an elaborate conspiracy to coerce Netflix into direct interconnect agreements, when they can thwart Netflix by simply continuing to slow its traffic?

The Slow Lane Blackmail version - Conclusion #

So overall, the evidence for the Slow Lane Blackmail version is spotty, and difficult to square with what actually happened. It has holes which can be plugged, perhaps persuasively, but not with facts, only with more supposition, inference, and conjecture. That doesn’t mean it won’t turn out to be true; it just means that given how little we know, it’s not a very convincing explanation of the Netflix direct interconnect agreements with ISPs.

Would OIO+c make Verizon’s refusal to upgrade legal? Since in the Slow Lane Blackmail version of the story Verizon isn’t charging anyone a fee — they are only refusing to act — “commercially reasonable” wouldn’t even apply.

Theory 4 of 6: The Fast Lane Blackmail version #

This version of the Netflix direct interconnect agreements attempts to get around one of the Slow Lane Blackmail’s problems: that ISPs had no incentive to coerce Netflix into a direct interconnect agreement, because just slowing Netflix backbone traffic already accomplished their goal.

According to the Fast Lane Blackmail version, direct interconnect is faster than going through a backbone, and that this fact provided ISPs with a perfect excuse to charge Netflix high fees for it. So they slowed the backbone connections, told Netflix that the problem was that backbone connections are just slower, and then overcharged Netflix for direct interconnect.

This is all conjecture; again, there is no solid evidence. It also makes no engineering sense, and it doesn’t fit the known facts.

A direct interconnect connection is not necessarily faster than going through a backbone. There are many reasons it can be slower, some of them quite complex; but here is a simple example. If Netflix purchases only a single direct interconnect connection to Comcast, say in Seattle, Washington, this would be wonderful for Seattle Comcast customers, but it might not help me very much 3,000 miles in Cambridge, MA, or any other Comcast subscribers on the East Coast. Not only is 3,000 miles a long distance for packets to travel, they have to “hop” through many Comcast computers on the way, and may get stuck in traffic jams jostling against all the other Netflix packets trying to get to the East Coast. If Netflix instead uses a backbone that connects to Comcast at multiple points around the north American continent, that will be faster for most Comcast customers. It’s like the difference between straight driving a shorter distance through a possibly crowded town, or taking a longer but possibly faster detour around it.

This isn’t just network theory; in fact, Verizon customers experienced worse Netflix traffic after the Netflix/Verizon direct interconnect deal, reportedly because they didn’t work out the technical details in advance the way Netflix had done with Comcast.⁶⁵

More importantly, why should the ISPs jump through all these complicated hoops to charge Netflix a fee, when they can just charge it to the backbone providers? ISPs shouldn’t care where they get the fee from, as long as they get it. The alleged reason, that direct interconnect is faster than a backbone’s paid peering connections, and therefore an ISP can charge more for it, doesn’t make sense, because it isn’t true. It implausibly assumes that Netflix and the backbone providers are fools who can have the wool pulled over their eyes easily, would require a lot of extra effort by the ISPs, and of course direct interconnects are not necessarily faster. Not only is it just another theory without proof, it isn’t a very good theory.

Theory 5 of 6: The Monopoly version #

In this version, the alleged problem is that Comcast has a monopoly: nobody can offer a faster connection to Comcast than Comcast itself! So that is why Comcast and the other giant ISPs want to blackmail Netflix into direct interconnect. It will always cost a lot, because there can never be any competition to push down the price; and so startups will never be able to afford it.

This argument is unconvincing because a startup would never want such a direct interconnect “fast lane” in the first place. A direct interconnect connection is like a pipe, and the packets that flow through it are like water. When Netflix needs to push more packets through the pipe than can fit, people’s movies start pausing: only some people get their movie packets on time. What Netflix needs in this case is not a “faster” pipe, but a wider one. Now let’s say OIO+c or some other law forced all Internet service providers to offer everyone as wide a pipe as Netflix’s. Would this help Netflix’s startup competitors? No, not a bit. A movie startup is only going to have a trickle of subscribers. Whether it trickles through a small pipe or large pipe makes no difference: the trickle will be equally fast in both cases.

The Monopoly version also doesn’t explain why Comcast would prefer a direct interconnect deal: Comcast has no competition for Comcast connections when it comes to backbones as well, for the same reason. So according to this argument, all paid peering deals with an Internet service provider are bad for the virtuous circle, because there is no competition to keep them low, whether they are direct or indirect. This may be true — it just isn’t an argument against direct interconnect or OIO+c. Maybe we need to worry about ISPs like Comcast getting so big that they can starting charging exorbitantly high paid peering fees, to make all their money off of rich Internet companies and shut out the poor ones. As far as I know, however, nobody has made this argument, let alone presented evidence to the FCC that it is actually happening.

We have now looked at five leading theories of what is behind the 2014 Netflix direct interconnect deals with Comcast, Verizon, and AT&T. Every one of them is based on circumstantial evidence, and every one of them is unconvincing.

There is now one theory left.

Theory 6 of 6: The Bad Backbone version #

The final version of the Netflix/Comcast deal I am going to present is perhaps the most plausible, although the evidence for it is far from solid. What’s more, according to this version, the Netflix/Comcast deal has nothing to do with net neutrality at all, or with OIO+c.

There have been two reports that in 2013, Netflix used a troublesome backbone provider named Cogent for transit to Verizon and Comcast, and that Cogent was responsible for Verizon’s and Comcast’s customers’ Netflix problems.

According to the reports, Verizon and Comcast accused Cogent of having promised Netflix to deliver more traffic than was in Cogent’s existing paid peering agreement with them. Cogent tried to argue that Verizon and Comcast ought to accept more data for free, because of Verizon’s and Comcast’s responsibility to their home Internet subscribers. Verizon and Comcast retorted that Cogent was simply trying to get a better paid peering agreement without paying for it.

In June, 2013, Arstechnica.com⁶⁶,⁶⁷ reported such a dispute between Cogent and Verizon; in February, 2014, streamingmediablog.com⁶⁸ reported the same kind of dispute between Cogent and Comcast. Neither report mentioned an additional “access fee”; they only reported paid peering disagreements. The streamingmediablog.com report says that Cogent was the last straw for Netflix, that it had finally had enough of using backbones, and that this motivated them to approach Comcast about a direct connection.

The Ars Technica reporters got official statements from both Cogent and Verizon that backed up their story: Cogent seemed to be arguing, bizarrely, that Verizon ought to give them extra bandwidth for free. The reporter of the Cogent/Comcast dispute didn’t want to reveal his confidential sources or their supporting data, which was proprietary.

If this is what happened, then Comcast didn’t blackmail Netflix at all; Netflix just decided it preferred a direct connection to dealing with middlemen. It seems plausible, perhaps more plausible than any of the other versions, but it too has problems. First, not all of the evidence is publicly available. Second, it doesn’t explain why Reed Hastings almost immediately turned on Comcast. According to this version, both Hastings and Comcast should have been content with their deal.⁶⁹

So was OIO+c a good idea? #

The FCC’s Open Internet Order was designed to protect the virtuous circle by making net neutrality the law. After a federal court struck down the Open Internet Order (OIO) regulation in January, 2014, the FCC proposed in May to salvage it by allowing “commercially reasonable” exceptions to the Order’s no-blocking and no-discrimination rules. This was a revision of the Open Internet Order which I have called OIO+c. OIO+c proposes a distinction between permissible and impermissible (reasonable and unreasonable) commercial behavior that explicitly outlaws any commercial behavior detrimental to the virtuous circle.

Two kinds of threats to the virtuous circle have been proposed that would allegedly be allowed (“commercially reasonable”) under OIO+c; if those two threats are convincing and the OIO+c cannot stop them, that would be a good reason to reject OIO+c.

We have seen that the first kind of threat, subsidized unlimited streaming, is real, but that the second kind of threat, so-called “fast lanes”, is most likely illusory.

That leaves only subsidized unlimited streaming. On its face, OIO+c appears to make it illegal; but that doesn’t mean the courts would agree. It all comes down to what lawyers and judges decide “commercially reasonable” means in this context. With OIO+c, the FCC’s heart is in the right place; the real concern is whether OIO+c would work.

But vulgar net neutrality has ensured that this will not be the decisive issue. Vulgar net neutrality portrays the issue as an evil plot by industry to manipulate the FCC into legalizing fast lanes that will destroy the Internet. Thanks to vulgar net neutrality, the public backlash against OIO+c has been enormous, and OIO+c will never have its day in court.

FCC Options #3 and #4: The Title II and Title II “lite” choices #

I need to amend our simple story. It should really be: a government agency tried to make something the law, failed, and is now trying again - again. What I never told you is that this story is a rerun. If we google “Title II FCC”, the results from 2010 are indistinguishable from those of 2014. The arguments are all the same, and so is the issue: whether the FCC should reclassify Internet service providers from “information services” under Title I, Section 706, to common carrier “telecommunications carriers” under Title II of the Telecommunications Act of 1934.

In early 2010, a federal court ruled that the FCC did not have the authority to stop Comcast blocking some of its Internet users’ Bittorrent downloads. For the rest of the year a battle raged in the media and Washington corridors over whether the FCC ought to reclaim this power by reclassifying all ISPs as common carriers under Title II. At the end of 2010, the FCC issued the Open Internet Order instead. We all know how that ended; and so here we are again, four years later, repeating the same tired arguments over Title II.

Those against Title II reclassification say it would impose onerous and expensive regulation that would increase consumer prices and drive smaller ISPs out of business; that it would allow the FCC to set those prices, eliminating ISP competition and so smothering innovation; and whether the courts ultimately allowed Title II reclassification or not, years of lawsuits would consume the FCC’s few resources and plunge the industry into ruinous legal uncertainty.

Those in favor of Title II argue that Internet access is now as essential as water, electricity, and the telephone to modern life, and should be regulated under common carriage rules as they are; that the dangers of regulation are grossly exaggerated, and anyway can be addressed with what I have called “Title II lite” (reclassifying, but promising to enforce only a few Title II regulations on ISPs); that we already subsidize underserved rural telephone networks by taxing urban ones, and can do the same for small rural ISPs; and that there is now, or will soon be, enough support in Congress and throughout the country for net neutrality to cow legal and political opposition from industry.

These are all legal, political, and economic arguments. As an engineer, I have little to add beyond the observation that as with the FCC’s other options, doing nothing or OIO+c, what both sides and the press present as overwhelming moral arguments are in fact arguments supported not by overwhelming evidence, but by evidence that is overwhelmingly circumstantial. This is no-one’s fault. Net neutrality policy is yet another issue that, despite our technological and scientific triumphs, has no clear resolution. The real sin of vulgar net neutrality is to make us believe otherwise, to divide us, and to destroy the least that we might hope for instead: reasonable compromise in the face of the humanly undecidable.

Four Lessons #

I do not know the extent of my ignorance on net neutrality. There is always more evidence, more history, another document with another revelation that adds depth to what I thought I already understood, or forces me to reconsider. I have changed so many high horses in midstream so many times, that I now instinctively distrust all confident pronouncements on net neutrality, as well as anyone who makes them, including myself. To my mind what this shows is not that net neutrality is a fertile subject, but a sterile one. There are no real, solid problems it would solve, and so everyone argues endlessly over the few facts that are publicly known.

Looking back now, I think there are four lessons we may draw from all this. Two are about net neutrality, and two are about how it has been covered.

Two net neutrality lessons #

The net neutrality debate has almost inadvertently exposed two very real problems for the virtuous circle that are rarely noticed.

The first is the perverse incentive that large Internet service providers have to discourage actual use of the high-speed Internet connections they sell. As we have seen, the more successful streaming media companies like Netflix are, the less customers of ISPs like Comcast and Verizon will want to buy Comcast’s and Verizon’s cable TV on-demand shows and channels. These ISPs have a strong incentive to neglect, if not discourage, their fast Internet connections. Those fast connections help Netflix sell Netflix streaming movies, to the detriment of Comcast selling Comcast streaming movies.

In other words, Comcast would prefer a world in which customers only buy Comcast streaming movies. In that world, customers wouldn’t need high speed internet to watch Comcast movies, because the movies all originate inside Comcast’s network. So as long as Comcast customers prefer Netflix streaming to Comcast’s streaming movies, Comcast will have an incentive to discourage sales of Comcast high speed internet to Comcast customers. The same is true of Comcast’s competitor ISPs like Verizon.

This perverse incentive only exists for an ISP that sells both streaming movies and cable TV, as well as Internet access. And that suggests an obvious solution: Congress or the FCC should break all these companies in two, and make it illegal for any company to sell both. Either you sell Internet access as an ISP, or you sell streaming movies and TV: but not both.

Think of it as the Internet’s equivalent of the Glass-Steagall Act, which forbade humble deposit-account banks from becoming high-risk investment banks, and vice versa. This would also encourage competition. Today, no startup ISP can make inroads unless it also sells cable TV. Only a giant like Google can do that, as it does with Google Fiber, its ongoing program to break into the ISP business. (And Google apparently gave up on – “paused” – Google Fiber in 2015.)

The second lesson is that the FCC or Congress ought to require all ISPs to make their Internet cables available to each other, at “commercially reasonable” rates.

During the previous two decades, companies like Verizon invested huge sums laying fiber optic and other high-speed cables throughout the United States. Digging trenches for cable is extraordinarily expensive, especially in cities. These companies took a huge risk that not enough people would sign up to recoup that cost; often they only laid the cable if a city or county agreed to hedge this risk with subsidies. The risk has since paid off, but at the time nobody could have known that it would.

A consequence of this history is that these companies will not share these lines with each other. A would-be startup ISP must lay its own cable all over again; once more, only a giant like Google can afford to do that. If companies were forced by Congress or the FCC to share their cables at a reasonable cost, we would finally have real competition among ISPs in this country. Reclassification under Title II would allow this, but there is no technical reason Congress or the FCC cannot legislate on just this narrow issue. The healthy competition it would create for reliable Internet and customer care would be amply justified under the Communications Act of 1934 alone, which created the FCC to foster world-class, world-wide, affordable and accessible information and wire services for all Americans.

Two vulgar net neutrality lessons #

There are two lessons we can draw from the damage vulgar net neutrality has done.

The first is for us as readers. The lesson here is to realize that journalists, bloggers, and activists, no matter how upstanding, virtuous, and well-intentioned, are under pressure to interpret complex issues in ways that generate more readers, more activism, and more donations, in a vicious circle that feeds upon itself. Vulgar net neutrality should teach us to be suspicious of any technology stories that arouse moral indignation and outrage.

The second lesson is for the media. News organizations that can afford it have a dedicated science “beat” that demands specialized knowledge, an ability to explain it engagingly, and years if not decades to earn the trust of knowledgeable and reliable sources. Without a long-term investment in an “engineering beat”, technology will continue to be covered merely as breathless announcements of amazing new gadgets and software, whose inner workings are never explained. This change cannot happen overnight, but until it does, we as a society will continue to make important decisions based on views that are not wrong, or ill-informed, or divisive, or popular, but all of these these things; views which, in a word, are vulgar.

Conclusion #

In politics more than anywhere else, everything begins with moral indignation. — Milovan Djilas

In 2014, the arcane subject of Internet “net neutrality” became another politically polarizing issue that divides Americans, like gun control or illegal immigration. Too few people understood how the Internet actually works, and belief in technical absurdities gave rise to views and moral indignation that have fueled a political movement. I refer to all these views as “vulgar net neutrality”. It is meant to be a label that is overly broad but still useful, like “liberal” and “conservative”.

Vulgar net neutrality claims to be in favor of net neutrality, and vulgar anti-net neutrality to be opposed to it; but both are just confused. Each acknowledges that Tim Wu introduced the term “net neutrality” in 2003, but in the same breath they describe it as a principle of treating all Internet data equally. Tim Wu actually favored discriminating between types of data.

Vulgar net neutrality has been like an autoimmune disease, turning blindly on its own main defender, the Federal Communications Commission. For years, the FCC tried to make net neutrality a philosophical principle of law, but not actual law: Michael Powell’s four “guiding principles.” When that failed, the FCC tried to make net neutrality itself the law, by making regulations for it.

When that failed in January, 2014, after a federal court ruling in favor of Verizon’s challenge to the regulations, the FCC tried to outwit the court in May, 2014. The FCC proposed weakening its net neutrality regulation, by allowing conduct the court had demanded – but only when that conduct did not violate what the original regulation was designed to protect.

This was a subtle and desperate legal gambit by the FCC, with two objectives: to save what vulgar net neutrality protesters immediately accused the FCC of trying to destroy, namely net neutrality on the Internet; and to oppose the corporations seeking to overturn net neutrality, whom the protesters accused the FCC of colluding with.

Legally and politically, the FCC’s May proposal was a calculated risk; as marketing, it proved a catastrophe.

Vulgar net neutrality triumphed; the FCC’s May proposal is now moot. But when we examine all the options the FCC faced, we find not only that none were promising, but that the alleged threats to the Internet are indistinct. The arguments by vulgar net neutrality proponents all turn out to be overwhelmingly circumstantial, and those proposed by the FCC itself are few in number and not very strong. My conclusion is that this is one reason we should dismiss the entire net neutrality debate as sterile, at least for now. The other reason is that what the debate does reveal are dangers to the Internet that have nothing to do with net neutrality: that ISPs ought not to be allowed to sell streaming entertainment and Internet access, and that ISPs ought to be required to allow “commercially reasonable” use of their fiber by anyone. We should all refocus our political energies on those two issues instead.

One striking aspect of the plague of the Middle Ages is that even a rudimentary awareness of microbes and infection could have contained it. That took science another 400 years. A rudimentary understanding of the Internet’s engineering is enough to defang vulgar net neutrality, and to save us from potentially disastrous decisions. But there is no need to wait 400 years, or even one, for that knowledge. The Internet has become as essential to modern life as electricity or clean water. To protect it, we just need to understand it a little better. At a minimum, that means the journalists covering the Internet for us need to be good engineers.

Vulgar Net Neutrality by Richard Bondi is licensed under a Creative Commons Attribution-NoDerivatives 4.0 International License.