By Timothy B. Lee
When Mark Zuckerberg created Facebook in his Harvard dorm room, he didn’t need to ask Comcast, Verizon, or other internet service providers to add Facebook to their networks. He also didn’t have to pay these companies extra fees to ensure that Facebook would work as well as the websites of established companies. As soon as he created the Facebook website, it was automatically available from any internet-connected computer in the world.
This aspect of the internet is network neutrality. And a lot of network neutrality supporters fear it’s in danger.
President Obama pledged to support net neutrality on the campaign trail in 2007. But many Obama supporters have felt let down by Obama’s choice to lead the Federal Communications Commission, Tom Wheeler.
Last week Wheeler announced a new set of network neutrality regulations. The details haven’t been released yet, but press accounts indicate that Wheeler’s proposal will allow internet service providers to offer a “fast lane” for online services, a concept that’s anathema to network neutrality stalwarts.
Yet Wheeler’s decision to water down network neutrality regulations isn’t even the biggest threat to the open internet right now. The internet itself is changing in ways that threatens to make the conventional net neutrality debate almost irrelevant. In recent weeks, Netflix has agreed to pay first Comcast and then Verizon for private connections directly to their respective networks. Netflix signed these deals under protest, charging that it had been coerced to pay “tolls” just to deliver content to their own customers.
That might sound like a net neutrality violation, but the practice doesn’t actually run afoul of the network neutrality rules advocates have been pushing for the last decade. Those rules ban “fast lanes” for content that arrives over the internet backbone, the shared information super highway that carries the bulk of the internet traffic today. But what Netflix paid Comcast and Verizon for amounts to a new, private highway just for Netflix traffic. Conventional network neutrality rules don’t regulate this kind of deal.
These private connections are going to be increasingly important to the American internet in the coming years. That might force net neutrality proponents to go back to the drawing board. Otherwise they might win the battle for net neutrality and still lose the war for a level playing field on the internet.
The problem with fast lanes
It’s a typical weekday evening and you and your neighbors are all using the internet in various ways. You’re watching Netflix videos, playing World of Warcraft, checking email, downloading podcasts, and reading cardstacks on Vox.com. The information required to display all this content is sent from servers all over the world. But it quickly finds its way to your internet service provider, the company that provides you and your neighbors with home internet access.
Internet usage is particularly heavy this evening, and your ISP doesn’t have the capacity to handle all the data you and your neighbors are downloading. So your neighbor’s World of Warcraft game starts to stutter. Another’s House of Cards episode freezes up and starts buffering. Your Skype video chat to your sister becomes pixelated and jerky. A digital traffic jam is ruining everyone’s internet experience.
Some people think a “fast lane” arrangement is the solution to this problem. Some applications are more affected by congestion than others. Some applications are more valuable to users than others. So maybe the network ought to give top priority to applications that need it the most. And why shouldn’t your ISP use the same method as FedEx to decide who gets the fastest delivery? Applications that need faster delivery can pay extra for it. The network might look like this:
Here, MyFlix has paid AT&Tcast to give its content priority over the content of its competitors. MyFlix customers get an excellent experience, but using YouBook or FaceTube might not be as pleasant.
But this isn’t how the internet works right now. For the most part, internet connections work on a first-come, first-served basis, with no one’s packets getting special treatment. And net neutrality supporters think that’s a good thing.
There are several arguments for this neutral internet model. One is simplicity. There are thousands of networks around the world. The miracle of the internet is that anyone can set up a web server, anywhere in the world, and instantly reach everyone else, no matter where they are or what network they’re using. But if broadband providers started dividing their networks up into fast lanes and slow lanes, things could get more complicated. To get satisfactory service for your website, you might have to negotiate fast-lane agreements with thousands of ISPs all over the world. Companies that didn’t have the money — or the manpower — to do that would be at a competitive disadvantage.
There’s also a danger that large internet service providers will abuse their monopoly power. Most of the leading American broadband companies also sell paid television services that compete directly with online streaming services such as Netflix and Amazon Instant Video. Network owners might be tempted to relegate online video services to the slow lane to prevent them from becoming a competitive threat to their lucrative paid television businesses. Or they might charge competing services a big markup for access to the fast lane, ensuring that they won’t be able to undercut them on price.
A final problem is that a multi-tiered business model could give ISPs perverse incentives. An ISP might be tempted to make its slow lane slower — or at least not upgrade it very quickly — to encourage content companies to pony up for fast-lane status.
At root, all of these arguments are about ensuring that the internet remains a fertile ground for new innovations. When Steve Chen, Chad Hurley, and Jawed Karim invented YouTube in 2005, they didn’t have to negotiate special fast-lane contracts with ISPs around the world. They also didn’t have to worry that incumbent broadband providers would view them as a threat to their cable services and relegate them to the slow lane — or demand fast-lane fees so high they couldn’t afford to pay them. YouTube could compete with much larger companies on a level playing field. Network neutrality advocates want to make sure it stays that way.
The changing internet
When an ISP receives the bulk of its traffic through one big transit provider, as in the figure above, network neutrality is relatively easy to define. It just means that the ISP needs to handle the packets it receives over that big pipe on a first-come, first-served basis.
That’s how things worked when Tim Wu, an academic who’s now a law professor at Columbia, coined the term network neutrality in 2002. ISPs purchased connectivity from a handful of companies that operated long-distance networks known as the internet’s backbone. Companies that provided this service, known in industry lingo as “transit,” acted as middle-men, carrying data between ISPs and content providers.
But the internet’s structure is changing. Both residential ISPs and content companies have been growing larger and more sophisticated. And increasingly, they are cutting out the middle-men. Instead of relying on transit providers to carry traffic between them, they’re connecting to each other directly. As a result, the internet increasingly looks like this:
The similarity between this figure and the previous one should be obvious. In Figure 2, MyFlix routed traffic through Global Transit (a transit provider), but it paid AT&Tcast to give its packets top priority. In Figure 3, MyFlix and AT&Tcast have established a direct connection, bypassing Global Transit altogether.
The practical result of the second scenario is identical to that of the first. MyFlix gets a fast lane, or more precisely a private “fast road.” As a result, its content reaches customers faster than content sent by FaceTube and YouBook. It’s worth asking: should net neutrality advocates be as worried about this kind of arrangement as they are about the “fast lane” in Figure 2?
In the fall of 2013, Comcast customers started to notice that their Netflix streaming experience was getting worse. The average video quality got lower and it became more common for videos to suddenly stop for “buffering.” These problems were cropping up because the bandwidth available to each Netflix subscriber on Comcast’s network was falling rapidly:
According to Netflix, this was happening because “Comcast allowed its links to internet transit providers like Level 3, XO, Cogent and Tata to clog up.” In other words, the big highways connecting Comcast with the rest of the internet were getting congested, and Comcast wasn’t investing in new lanes. Comcast wouldn’t speak to me on the record, but in a blog post the company doesn’t dispute that clogged transit connections were responsible for the declining performance; it just blames Netflix for the problems.
Comcast and Netflix were playing a high-stakes game of chicken. Netflix, and the companies Netflix had paid to deliver its content, thought Comcast should accept traffic from them for free. Comcast thought it should be paid to accept the traffic. The result was a months-long standoff. Netflix usage increased, but the network didn’t get upgraded to accommodate the traffic growth. So each user got less and less bandwidth.
In February, Netflix blinked. Unable to get the speeds it wanted through third-party transit providers, it agreed to pay Comcast for a direct connection between their networks. Here’s what happened next:
Comcast says this was an ordinary business negotiation. Netflix says Comcast was making unreasonable, even unprecedented, demands. But Netflix lost the negotiations — of course that’s what they would say.
This might seem like a classic he-said, she-said situation, but Netflix has a point. Industry insiders wouldn’t talk to me on the record, but sources who have been on both sides of this kind of negotiation tell me Comcast’s demands were highly unusual. Normally, broadband ISPs pay transit providers like Level 3 and Cogent to deliver data their customers have requested. Here, Comcast was demanding that money flow in the opposite direction.
In a competitive broadband market, Comcast’s gambit wouldn’t have worked. “Here in Europe with a lot of competitive networks (to each home) that’s a difficult strategy to follow,” says Rudolf van der Berg, a telecom analyst at the Organization for Economic Cooperation and Development, based in Paris. “You’re punishing your customers and they can go to another operator relatively easily.”
But most Comcast customers have no more than one alternative. And for many of them (including people like me who are only served by Verizon DSL) that alternative is so much slower than Comcast’s network that it’s not a serious option. So Comcast didn’t have to worry about customer defections the way a European ISP would. That means it wasn’t taking much risk by refusing to upgrade its transit links.
In contrast, Netflix does face serious competition. Services like Amazon Instant Video, Hulu, and iTunes would love to lure customers away from Netflix. Customers can also get video content from YouTube, broadcast television, and many other sources. So Netflix’s declining video quality was inevitably costing it customers.
Comcast’s victory in its dispute with Netflix didn’t surprise Vishal Misra, a computer scientist at Columbia who has modeled this kind of negotiation. “If there is competition on the content side but not on the eyeball side that increases the leverage that the eyeball side has on the content side,” he wrote in a February blog post. If customers who subscribe to both Comcast and Netflix are dissatisfied, they can dump Netflix a lot more easily than they can dump Comcast. That gave Comcast the upper hand.
A broader concept of network neutrality?
We shouldn’t be too concerned for Netflix. It’s a big company, and the fees Comcast is demanding won’t drive them into bankruptcy. The larger problem is that this strategy of delaying network upgrades to force Netflix to the bargaining table creates a lot of collateral damage.
Comcast says it wasn’t deliberately degrading the performance of Netflix traffic, which would have been a violation of network neutrality. But that implies that every website that used the same transit provider as Netflix was experiencing similar performance problems. That’s obviously bad for Comcast customers, and it’s also bad for startups that are trying to become the next YouTube or Netflix. If the only way to get excellent service on America’s largest broadband networks is to negotiate a private connection directly to those networks, smaller companies with less cash and fewer lawyers are going to be at a competitive disadvantage.
In other words, those public transit links that were providing Netflix with subpar service could become the de facto slow lanes on Comcast’s network, while private, direct connections could become the fast lane.
Yet interconnection disputes are generally treated as a distinct issue from network neutrality. Neither the FCC’s 2010 Open Internet Order, which the courts struck down earlier this year, nor FCC Chairman Tom Wheeler’s new proposal, would govern this kind of interconnection dispute. Wheeler has said that he intends to monitor the interconnection situation closely, but he hasn’t put forward any concrete proposals to address the issue.
One reason for this is that Comcast’s hardball tactics are relatively new. According to Wu, “we never thought much about” interconnection issues when the concept of network neutrality was developed in 2002. That was “mainly because it seemed like, ‘Why would anyone do that?'” It was generally assumed that consumer ISPs would always buy enough transit to serve their customers’ needs.
Wu worked in Silicon Valley prior to 2002. And he says he remembers that “solving a joint technical challenge” — that is, making the internet work as well as possible — “was the predominant motivation of the engineers.” That’s still true in most parts of the internet. But some of the largest ISPs now seem to view declining network performance not as a technical problem to be solved so much as a source of leverage in business negotiations.
Another reason is that regulating interconnection is much more complex than a “classic” network neutrality rule. When all of an ISP’s traffic comes through one cable, it’s not too hard to write a rule requiring that the packets in that cable be treated equally. But it’s harder to write a rule governing when and how ISPs must interconnect. Someone needs to pay for the cost of these connections, and the fairest way to split the costs depends on many subtle factors, including geography, traffic patterns, and the relative size of the interconnecting networks. A poorly written interconnection rule could create a lot of work for lawyers without actually preventing abusive practices.
A good first step would be to require more transparency. “The big challenge with this whole internet interconnection world is that everything is opaque,” says Kevin Werbach, a scholar at the University of Pennsylvania who has long emphasized the importance of interconnection in broadband regulation. “You’re seeing this now with Netflix and Comcast, both sides are firing back and forth making representations about what’s going on. The public doesn’t actually know.”
Wu says he’s still mulling the best way for regulators to address this new challenge. But he believes one promising approach would be for regulators to mandate that large ISPs accept traffic bound for their own customers without payment.
Concerns about abusive interconnection practices also provide an argument against Comcast’s proposed merger with Time-Warner. Comcast’s sheer size gives the company leverage in its negotiations with both content and backbone companies. Further consolidation could make the problem even worse.
The bottom line is that network neutrality advocates will need to broaden their thinking to respond effectively to the internet’s changing structure. Merely banning fast lanes isn’t going to accomplish much if the largest ISPs are allowed to sell new private roads.