April 2008

Comcast recently unveiled a 50 Mbps broadband connection for $150/month, and has promised to have it available nationwide (and most likely bring the price down somewhat) by 2010. Verizon is in the process of rolling out a new fiber infrastructure that will allow it to offer a similar deal (you can also get an 8 Mbps connection for $50/month). All of this makes Qwest look like a relative laggard, with its announcement that it will soon be offering a 20 Mbit service for $100/month, and 12 Mbps for $50/month. And AT&T brings up the rear with plans for “only” 10 Mbit service.

One sometimes sees hand-wringing about the anemic state of the American broadband market. To the extent that other countries are doing even better (a debatable contention), we should certainly be looking for ways to make the broadband market more competitive. No doubt things would be progressing even faster if there were more players in the market. But the strong claim you sometimes see that there’s something deeply dysfunctional about the progress of the US broadband industry, is positively silly.

If Comcast and Verizon deliver on their promises to roll out 50 Mbit service by the end of the decade, and if prices follow their historical pattern of dropping over time, consumers in their service footprints will have seen the average speed of a $50 Internet connection increase by three orders of magnitude in about 15 years. And it’s been a reasonably steady pace of growth: in round numbers, $50 would buy you a 56kbps connection in 1998, a 512kbps connection in 2001, a 5Mbps connection in 2006, and (I hope) a 50 Mbps connection sometime early next decade. Things have consistently improved by a factor of 10 every 4-6 years.

It’s interesting to look back at the broadband argument we were having a couple of years ago. In mid-2006, TLF reader Luis Villa was reporting that he and his folks were still seeing typical broadband speeds of 2 Mpbs. Maybe he can chime in to let us know if that number has gone up at all. Personally, my apartment in St. Louis gets around 5 Mbps for about $30/month, and the house where I’m currently staying in DC has a 17 Mbps Comcast connection.

One of the things that makes it hard to judge is that broadband speeds and prices don’t tend to grow continuously and evenly around the country. Rather, carriers take turns leap-frogging one another, with each upgrade accompanied by a temporary price increase. So it can be tricky to judge the average rate of improvement by looking at just one market, because one market may seem to stagnate for several years at a time. But if one looks at the country as a whole, and focuses on time horizons closer to a decade, I think it’s undeniable that things are improving at a fairly rapid pace.

SHARE: 6 comments

Don Marti offers a good example of the point of my last post:

The Smoot-Hawley Tariff was law. Many of the “code is law” design choices are similar: they can cause big effects without necessarily implementing the lawmaker’s intent.

I couldn’t have put it better myself. And this brings to mind another important point about “law” that I think is missed in Lessig’s “code is law” formulation: the real “law” Lessig is analogizing computer code to often isn’t exactly omnipotent. The War on Drugs is law but there’s still lots of drug use going on. Strict gun control is law in some big cities but there’s still plenty of gun violence. The McCain-Feingold soft money ban is law but there’s still plenty of corruption in Washington. The copyright system is law but there’s still plenty of illicit file sharing going on. And in all of these cases, the law has had significant effects that were very different from their authors’ intent.

And indeed, if we’re going to analogize ISPs to legislators, it’s important to remember that legislators have several key advantages over ISPs. In the first place, actual legislators can have you thrown in jail, whereas the worst an ISP can do to you is disconnect your service. Second, governments have effectively unlimited sums of money to waste on futile efforts like the drug war. In contrast, ISPs are profit-seeking institutions so there’s some limits on the amount of idiocy they can undertake before they’re forced to stop due to financial considerations. Finally, the law has at least some moral suasion behind it; the fact that something is illegal at least causes people to mostly hide the fact that they’re doing it. In contrast, people evading the “law” of an ISP would be able to claim the moral high ground.

Yet despite all these advantages, our government hasn’t come close to stopping people from using drugs, sharing files, hiring prostitutes, owning guns, corrupting public officials, and so forth. Not because the laws for those purposes were badly designed, or because the people implementing them haven’t tried hard enough, but because there are fundamental limits to government’s ability to control the behavior of citizens in a free society.

I think something similar is true of a large-scale network like the Internet. Incompetent network owners, as the “governments” of their corners of the Internet, can certainly cause a lot of havoc, just like real governments. But their ability to control what their customers do is constrained for many of the same reasons that governments’ ability to stop drug use or file sharing is constrained. And that means that network owners that try to restrict their customers’ use of their network are as likely to shoot themselves in the foot as they are to enhance their bottom lines.

SHARE: 10 comments

Writing my critique of Larry Lessig’s Stanford lecture brought to mind an important ambiguity in Lessig’s oft-repeated slogan that code is law. I think the phrase actually has two distinct meanings that are often conflated but actually have quite different implications. One is, I think, indisputably true, whereas the other one is at best greatly overstated.

It’s certainly true that code is law in the sense that technological architectures place important constraints on the actions of their users. A TCP/IP network allows me to do different things than Verizon’s mobile market or the global ATM network. Wikipedia is the way it is largely because the software on which it runs constrains users in certain ways and empowers them in others. CALEA likely has given law enforcement agencies surveillance powers they couldn’t have had otherwise. To the extent that this was Lessig’s point, he’s obviously right.

However, when Lessig says “code is law,” he often seems to be making a significantly stronger claim about the power of code as law: not simply that code constrains us in some ways, but that the authors of code have a great deal of control over the exact nature of the constraints the technology they build will place on people. So that virtually any outcome the code designer wants to achieve, within reason, can be achieved if the code is designed by a sufficiently clever and determined programmer.

This stronger formulation strikes me as obviously wrong. This is true for at least two reasons. First, the set of tools available to the code-writer is often rather limited. For one thing, barring major breakthroughs in AI technology, many concepts and categories that are common sense to human beings cannot easily be translated into code. Rules like “block messages critical of president Bush” or “don’t run applications that undermine our business model” can’t easily be translated into hardware or software. There are a variety of heuristics that can be used to approximate these results, but they’re almost always going to be possible for human beings to circumvent.

The deeper challenge is a Hayekian point about spontaneous order: with a sufficiently complex and powerful technological platform, it often will not be possible to even predict, much less control, how the technology will be used in practice. Complex technologies often exhibit emergent properties, with the whole exhibiting behaviors that “emerge” from the complex interplay of much simpler constituent parts. It would have been hard for anyone to predict, for example, that the simple rules of wiki software could form the basis for a million-entry encyclopedia. Indeed, it’s pretty clear that Wikipedia was possible only because the site’s creators gave up any semblance of centralized control and allowed spontaneous order to work its magic.

A similar point applies to the Internet, and to the network owners that nominally control it. They certainly have some levers they can push to change certain micro-level characteristics of their networks, just as Jimmy Wales could make changes to the code of Wikipedia. But there’s no reason to think that either Wikipedia or Comcast would have any way to either predict or control what effects any given micro-level change to their platforms would have on the macro-level behavior of the whole system. They’re both large, complex systems whose millions of participants are each pursuing their own idiosyncratic objectives, and they’re likely to react and interact in surprising ways.

Now, the fact that Comcast can’t predict or control what effects its tinkering with its network might have does not, of course, mean that it won’t try. But it does mean that we should be skeptical of just-so stories about telcos turning the Internet into a well-manicured walled garden. When you create complex, open systems, they tend to take on a life of their own, and once you’ve ceded centralized authority, it can be very difficult to take back. Code may be law, but I think it’s a much more limited kind of law than a lot of people seem to think.

SHARE: 7 comments

Over at Techdirt, I’ve got a series of posts that highlight some of the major arguments in my forthcoming paper on network neutrality. I’m particularly pleased with the latest installment of the series:

He claims that “owners have the power to change [the Internet’s architecture], using it as a tool, not to facilitate competition but to weaken competition.” Do they? He doesn’t spend any time explaining how networks would do this, or what kind of architectural changes he has in mind. But he does give an example that I think is quite illuminating, although not quite in the way he had in mind.

Lessig imagines a world of proprietary power outlets, in which the electricity grid determines the make and model of an appliance before deciding whether to supply it with power. So your power company might charge you one price for a Sony TV, another price for a Hitachi TV, and it might refuse to work at all with an RCA TV. Lessig is certainly right that that would be a bad way for the electricity grid to work, and it would certainly be a headache for everybody if things had been set up that way from the beginning. But the really interesting question is what a power company would have to do if it wanted to switch an existing electricity grid over to a discriminatory model. Because the AT&Ts and Comcasts of the world wouldn’t be starting from scratch; they’d be changing an existing, open network.

I focus on an aspect of the debate that I’ve seen receive almost no attention elsewhere: assuming that network discrimination remains legal, how much power could network owners actually exert over the use of their networks? There’s an assumption on both sides that ownership of a network automatically translates to comprehensive control over how it’s used. But I have yet to see anyone give a plausible explanation of how a last mile provider would get from the open network they’ve got now to the tightly-controlled one that network neutrality advocates fear. It’s always simply asserted, as Lessig does here, and then taken as a given for the rest of the argument. It’s a weakness in the pro-regulation argument that I wish more critics of regulation would highlight.

I go on to discuss the difficulties of real-world architectural transitions. Read the whole thing here.

SHARE: 6 comments