Whose Internet is it, anyway?

One of the Internet's chief architects looks at the FCC's proposed Net neutrality rules.

Last week, the new chairman of the Federal Communications Commission, Julius Genachowski, broke with precedent by proposing federal rules that enforce Net neutrality — the principle that Internet service providers (ISPs) shouldn't play favorites with the traffic traveling over their networks.

Proponents argue that Net neutrality promotes innovation. If software developers find more efficient ways to use the Internet, the argument goes, they shouldn't fear reprisal from ISPs that sell competing products. Broadband providers that also offer landline phone service shouldn't degrade the quality of Internet telephone calls in order to preserve their market share; the same goes for cable companies and Internet video.

But ISPs argue that they sometimes need to throttle back traffic sent by heavy users. Otherwise, they say, the network will become congested and slow to a crawl; thousands of casual users will pay the price for a few customers sucking up a disproportionate share of bandwidth. If they lose the ability to regulate traffic, the ISPs argue, they'll have to greatly increase network capacity — and their customers will bear the cost.

David Clark, a researcher in the Computer Science and Artificial Intelligence Laboratory who for most of the 1980s was the Internet's chief architect, has been following the Net neutrality debate for decades and spoke with the News Office about the FCC's proposed rules.

Q: In what respect do ISPs have a legitimate concern?

The Internet is not, in terms of cost for byte, terribly expensive, but neither is it free. You can find some numbers reported informally in the press, and I think the numbers are somewhat reasonable, that for a residential ISP to deliver another gigabyte of information to you, the cost in terms of the investment they make in facilities allocated to that gigabyte is about ten cents. So if I watch Internet television eight hours a day every day of the month, I'm probably generating several dollars in cost. It's not several hundred dollars in cost; it's several dollars in cost. But that's probably the ISP's profit margin.

Q: So what can the ISP do?

A while back Comcast announced that they were putting a monthly cap on their Internet users over the cable system. The cap they announced was 250 gigabytes a month. And nobody blinked, because that's maybe 50 or 100 times what the average Internet user was doing.

What if I said to you, okay, for $40 a month, which is what most people pay today, I'm going to do something much more restrictive than what Comcast did: you can transfer 100 gigabytes? For $50 a month, we'll take the cap off, and you can transfer as much as you want. For an additional $10, would the high-end guys be willing to do that? A lot of people today pay a little extra to get a higher peak rate; many people subscribe to a premium version of Internet service. I think most people would say, if the high-end people are paying an extra $10 a month, that's not burdensome.

People's fear in this space is that if we take one step away from the current pricing model of all-you-can-eat flat pricing, that the world will end. All of a sudden we'll be paying by the byte, which I think everyone understands will be a real inhibitor of experimentation on the market.

Q: But why is a usage cap any better than paying by the byte?

I was talking to somebody in a school district, and they said, look, we couldn't possibly afford a per-byte charge because some kid could come and get a program running on the computer and leave it running over the weekend and blow our entire year's budget.

I really think that's the point. The user at home wants to be protected from amazing overage charges. His computer goes into a loop, or it has a virus, and the computer has five days where it does nothing but splash data out full time, and you get a bill at the end of the month for $5,000. That's what terrifies everybody. But in the wireless space, many of the broadband services are fixed price with a usage cap, and the market deals with that much better than with a per-byte charge. Because nobody knows with an Internet application how many bytes it sends. Will this cost me a penny or a dime or a dollar? But they can average over a month. They look at the bill: I sent three gigabytes last month. The cap was five. Okay! They can deal with that.

The only question is, when usage caps come in, will they be done in a reasonable way, or will lack of competitive discipline allow ISPs to try things that are really pretty abusive?

Q: But given that many cable providers and phone companies are basically local monopolies, is there enough competition to provide that discipline?

As a rule of thumb, it's nice to see four or five competitors in a market. And we only have two wireline [phone and cable] in most markets. So you might say that two isn't really enough. On the other hand, when I watch Comcast and Verizon, in our serving area here, slugging it out on television with their ads, boy there's a lot of competition going on there. Just observing what I've seen on television, they believe that they're in a very competitive situation. Comcast just sent me a note that said, "We've upgraded your service." Why'd they do that? Because they're subjected to the pressures of competition.

Q: One of Chairman Genachowski's comments that's gotten a lot of attention is that Net neutrality rules will apply to wireless services as well. What do you see happening there?

Spectrum is more scarce than, say, the capacity on the fiber to your house. When you get into a heavily used cell where a number of people are trying to do bit-rate-intensive things, there are going to be real issues in managing that scarcity and allocating it. I quoted you a number of what it cost to do a gigabyte: that number applies to an Internet service provider that's large, that's got scale, and that's probably operating in a metropolitan or suburban area. People don't want to show you their exact business models, but I've seen situations that look like that number for a rural wireless provider was more like a dollar a gigabyte.

I think the thing we're going to debate in the wireless space is whether or not there are classes of behaviors that seem to be associated with classes of applications. Should those behaviors be limited? Whether the wireless guys will say, "Look, you just can't watch as much video as you want." And they can do that in two ways. One of them is, they can say you have a monthly cap of three gigabytes. Go crazy! You want to watch video, you can blow out your monthly quota in about two days. And then you're going to be cranky. Or they could say, we're going to block certain video applications. I'm in favor of a usage cap over application-specific discrimination. Because the usage cap really does reflect to some extent what the ISP's cost structure is. Give the consumer choice.

Topics: Internet, Computer Science and Artificial Intelligence Laboratory (CSAIL), Computer science and technology, Federal Communications Commission (FCC), Net Neutrality


Living in Australia has given me a much more tangible understanding of the issue of usage caps. It is something of great concern because their effective implementation demands a broader sense of perspective than just how much the market will bare.

As Dave notes, there are many plans here with monthly quotas that, if one isn't paying attention, or more accurately is using the network for the real value it can bring, will draw your quota down to zero in days. Dare to watch video on the net and your monthly allocation will disappear before your eyes. When that happens there are two kinds of plans - either you're network access will be throttled to 56kbps (dial-up speeds) until your monthly allocation is renewed in the following month; or, you'll find yourself paying what amounts to cell phone service international data roaming rates and your bank account will rapidly asymptote to zero. Either way, it's a punishment for use.

In the provider's eyes you're overusing the service. In the user's eyes you're trying to take advantage of the growing proliferation of media rich applications, and these demand bandwidth the providers haven't anticipated. Recall the 'surprise' that AT&T has expressed at finding mobile iPhone users consume 3x4 times the average data relative to other mobile users. This outcome for what was always presented as a mobile internet device has caught an ISP that should have known better. What does that suggest for ISPs? It makes me profoundly worried.

Like the adage describing the negotiation over an assignation in which one party is asked if they'll spend time for "x" dollars with the other and refuses, but agrees to for "x+n" dollars and agrees. The sage assessment came as "now that we've established what you are, we're simply haggling over price." This issue describes the worry.

Perhaps quotas with a waiver for areas identified as in the public interest might work - e.g, education, healthcare, etc. But I'd hate to be on the committee making those distinctions. I'd prefer that the flat rate be modeled such that the majority of users are in fact profitable to the ISP and only those in the marginal extremely high use category are singled out (e.g., maximum bandwidth utilized by a specific user continuously across 24 hour periods).

Gee Whiz expressed disappointment that the MIT News 'published an article' that didn't reflect what them claimed represented high "standard", or, "academic rigor". Hello??? This is an opinion piece be a person who is uniquely sitting with a vantage point to offer their personal insights based on a deeply engaged history with the development of the internet. It is NOT an academic article. It wasn't purported to be. I for one want to hear these opinions and recognize them as such. If Gee Whiz wants to read referred journals their looking in the wrong place. Any librarian would be happy to help them find the kinds of articles they must be looking for when they mistakenly thought the MIT News opinion piece was a referred publication. Apply the evaluation criteria appropriate to the publication rather than criticize it for what it's not nor ever suggested it was. It's a classic tactic but one that is not welcomed nor does it contribute to the conversation.

I was astounded to find this article on an academic website. The author notes that usually one would expect at least four or five members in a market to guarantee competition. His sole factual argument against this is "On the other hand, when I watch Comcast and Verizon, in our serving area here, slugging it out on television with their ads, boy there's a lot of competition going on there." Well, gee whiz! They argue a lot so there must be competition. Does anyone out there think this is a convincing argument? Wouldn't a more convincing argument be "We have some of the fastest internet access in the world"? Problem is, we don't. We have a ridiculously slow internet. Japanese users, for example, regularly get 100 MB/sec, where we often have to pay a PREMIUM for 10. Gee, maybe there isn't real competition at all! Maybe we are paying through the nose for poor access because of monopolies. How can an academic website publish an article like this with no standards whatsoever for accuracy or academic rigor. I am profoundly disappointed in MIT.


Thanks for reading my interview with Dr. Clark so attentively.

One of the (many) reasons for the relaunch of the MIT News site (less than two weeks ago) is that it's becoming more and more common for people to get their news online, from a number of different sources. We're hoping that readers interested in science and technology — whether inside or outside the Institute — will make the MIT News site one of their daily stops. To that end, we will (usually) be posting three new news stories a day. Like the science and technology stories you might find at other news outlets, ours will seek comment from experts outside MIT on the context and significance of the work they describe.

Also like other news outlets, we will sometimes reverse that formula, seeking comment from MIT experts on the news of the day. My interview with Dr. Clark was one such story. It is not a peer-reviewed paper in a journal; it is the type of interview you might find at the front of a news magazine. The point of such an interview is to contribute to the public discussion of an important topic by airing the particular views of a subject expert. In objecting to Dr. Clark's point about competition, you are taking part in that public discussion — something we very much encourage.

—Larry Hardesty

Back to the top