Death to Nuance!

Susan Crawford rings in the New Year in the Yale Law and Policy Review with an article (The Looming Cable Monopoly) that illustrates a prime reason that it’s nearly impossible to have a discussion about Internet policy in the United States without a food-fight breaking out. Crawford tells us, unequivocally, that a cable company takeover of the Internet is imminent, according to no less an authority than the National Broadband Plan. The Plan doesn’t say this, of course, but Crawford twists it into a pretzel to make it seem so. First she cites some text from the National Broadband Plan that speculates about one possible broadband future and finds a dark cloud around the silver lining of higher speeds:

Analysts project that within a few years, approximately 90% of the population is likely to have access to broadband networks capable of peak download speeds in excess of 50 Mbps as cable systems upgrade to DOCSIS 3.0.^[7] About 15% of the population is likely to be able to choose between two robust high-speed service services [sic]—cable with DOCSIS 3.0 and upgraded services from telephone companies offering fiber-to-the-premises (FTTP).^[8]

These upgrades represent a significant improvement to the U.S. broadband infrastructure, and consumers who value high download and upload speeds will benefit by having a service choice they did not have before the upgrade. The upgrades may, however, change competitive dynamics. Prior to cable’s DOCSIS 3.0 upgrade, more than 80% of the population could choose from two reasonably similar products ([Digital Subscriber Line (DSL)] and cable). Once the current round of upgrades is complete, consumers interested in only today’s typical peak speeds can, in principle, have the same choices available as they do today. Around 15% of the population will be able to choose from two providers for very high peak speeds (providers with FTTP and DOCSIS 3.0 infrastructure). However, providers offering fiber-to-the-node and then DSL from the node to the premises (FTTN), while potentially much faster than traditional DSL, may not be able to match the peak speeds offered by FTTP and DOCSIS 3.0. Thus, in areas that include 75% of the population, consumers will likely have only one service provider (cable companies with DOCSIS 3.0-enabled infrastructure) that can offer very high peak download speeds . . . .^[9]

[Ed: Emphasis added] Then she “translates” the Plan’s analysis into simplistic, Manichaean terms that the apparently mis-educated readers of the Yale Review require:

Here is a translation of this section: Where Verizon FiOS service exists, there will be competition with cable Internet access service providers for high-speed Internet access at speeds that are necessary to carry out real-time video conferencing or watch high-definition video. Where FiOS is not installed, there will not be any competition, and consumers will have just one provider to choose from: their local cable monopoly. Most Americans—perhaps as many as 85% of us—will fall into this latter category.

[Ed: Emphasis added] Isn’t that special? The Plan offers a hypothetical speculation about one possible broadband future, and Crawford neatly takes away all the subtlety and nuance and certifies this possibility as the One True Outcome, thus relieving the overburdened reader of any responsibility for thoughtful analysis and grappling with inconvenient facts. As networks get faster, choice will certainly evaporate. Worse yet, she puts words in the Plan’s mouth that it never spoke. Translation: We’re All Gonna Die! Gadget blogs such as Ars Technica were only too willing to treat Crawford’s hatchet-job on the National Broadband Plan with an authority it doesn’t deserve:

Crawford is a long-time advocate of net neutrality and related tech policy issues, so this might sound more than a bit hyperbolic to some observers. But one has only to listen to the ISPs themselves to see just how much they would love such a future (and it’s one of the reasons they fought so hard to limit FCC oversight of “specialized services” in the recent network neutrality rulemaking).

The Plan itself goes on to frame this speculation with the kind of humility that any exercise in futurism should employ:

As with fixed-mobile substitution, how the evolution of network capabilities affects competition depends on how pricing, consumer demand, technology and costs evolve over time. For example, if users continue to value primarily applications that do not require very high speeds (e.g., speeds in excess of 20 Mbps), and are not willing to pay much for vastly increased speeds, then a provider may not gain much of an advantage by offering those higher speeds. In contrast, if typical users require high speeds and only one provider can offer those speeds, and expected returns to telephone companies do not justify fiber upgrades, then users may face higher prices, fewer choices and less innovation.

Crawford’s certain future fuels her battle-cry for line-sharing, the same regime she was promoting long before the National Broadband Plan was a twinkle in Blair Levin’s eye:

We are about to confront a well-coordinated cabal of local monopoly cable providers. When it comes to affordability, ubiquity, and nondiscrimination, we could decide to take a lesson from a host of other developed nations—particularly Australia. As a report from the Berkman Center made clear earlier this year, policies requiring line-sharing at regulated rates have played a central role in the spread of low-priced, nondiscriminatory, very-high-speed access in many other nations.^[20]

The Australian network hasn’t even been built (and may never be,) so it’s hard to see exactly how it applies to the dilemma of our uncertain future; the Plan itself reached no such conclusion. Rather, it proposed a more rational response to its own speculation:

Because of this risk, it is crucial that the FCC track and compare the evolution of pricing in areas where two service providers offer very high peak speeds with pricing in areas where only one provider can offer very high peak speeds. The FCC should benchmark prices and services and include these in future reports on the state of broadband deployment.

Crawford wants to immediately impose line-sharing on the “well-coordinated cabal of local monopolies” but the Plan prefers to provide them the favor of trial before stringing them up.

The Plan carefully considered the pros and cons of line-sharing, and commissioned a report from the Berkman Center to answer the specific question that line-sharing raises: What price must be set for wholesale access to a network infrastructure such that both the wholesaler and the retailers have incentives to invest? Levin explained this in a recent back-and-forth on “Stupid Network” advocate David Isenberg’s blog:

I asked Yochai [Benkler, the Berkman Fellow who delivered the report,] to answer [a question that] he thought was irrelevant but I, having been through the ‘96 Act implementation, kinda thought was important: What is the wholesale discount necessary to incent the investment necessary for an unbundling strategy to work? ... What do you do if you have two facilities based carriers?

…

We could lower prices in a nanosecond by simply regulating them to be 99% lower. The question is always, what does that do to the world going forward….in a monopoly setting like Europe or Asia. In a duopoly setting it is a lot harder. And my point about the discount rate goes into a simple ROI analysis which suggests that it will prove pretty much impossible for anyone to invest into that model.

[Ed: Comment de-Blackberried]

The Berkman “report” dodged the question that Levin asked them to answer, and without an answer to that question all that line-sharing can ever be is a recipe for stagnation. Set the wholesale price too high, and there won’t be any independent ISPs; set it too low, and the copper wire to the home will never be replaced by fiber unless the government takes direct control of the cable plant, as they’re doing in Australia. Or as may happen in Australia, because their new network is barely off the drawing board, and unless the Labour Party stays in power it’s far from a sure thing. The Australian Conservatives are against it and consumers are lackluster; the initial deployment in Tasmania ran into the unexpected obstacle of 50% of consumers refusing to give permission to the government to give them a free hookup.

So we have more of the same, some vigorous hand-waving in favor of nationalizing communications infrastructure when most other countries are trying to privatize it in order to stimulate investment and competition, supported by distortion of the facts and irresponsible speculations about a future that’s far from certain.

Is it really a given that a cable monopoly over high-speed Internet services to the home is a “looming crisis” because DSL can’t compete and Verizon FiOS won’t compete, as Crawford eagerly declares and the credulous gadget blogs reinforce? Not really, for a number of reasons. Let’s consider an alternate future scenario.

First some facts: The gap between the modern vDSL+ systems deployed from Fiber to the Node (FTTN) systems by AT&T and Qwest and DOCSIS 3 isn’t as great as it appears.  FTTN runs at speeds from 25 Mbps up to 50 Mbps (the latter requires pair bonding) and will be faster in the future thanks to signal processing advances such as vectoring, a system that produces speeds up to 300 Mbps at 400 meters in lab tests meant to simulate real-world conditions using pair-bonding and the third “phantom pair” that vectoring enables.

For another, the speeds that DOCSIS 3 reaches – 160 Mbps with four channels bonded – are shared by a collection of 150-400 homes, while vDSL+ capacity is non-shared (or more correctly, only shared above the neighborhood node.) There are applications in which the high peak speeds of a shared network like DOCSIS are a huge advantage, such as web surfing and e-mail, but the applications Crawford cites (high-definition video streaming and conferencing) aren’t among them. For these applications to work well, we need networks with a lot of un-shared capacity, because they apply a persistent, constant load, not the kind of bursty, sporadic load that high-peak shared networks like best. Divide that 160 Mbps among 160 active home networks and you have a challenge that vDSL+ will happily confront. As more consumers shift their Internet usage from web surfing to Netflix-style video streaming and HD video conferencing, the value of networks that provide high peak experience at the expense of high persistent capacity diminishes.

When the telco “cabal” rolls out the next generation of DSL, the cable company “cabal” won’t sit back and do nothing, however. The persistent load capacity of cable networks can be increased by reducing the number of homes that share a DOCSIS channel group, first by Virtual Node Splits and then by physical node splits. So the population of roughly 300 homes per DOCSIS switch will come down to 150, 75, and then to a number close to the 32 that share a node in FiOS. The 160 Mbps that DOCSIS 3 offers with 4 bonded channels can double with currently installed gear, and then double again with new gear. Cable can potentially offer multiple gigabits/sec shared across an ever-smaller pool of users, just as Fiber to the Node can evolve to Fiber to the Home as the need develops.

As the capacity of DSL and cable networks is improving, video compression technology will continue to improve, which will multiply the effects of improved network capacity.  As these technologies develop, network operators will address the real bottlenecks for HD streaming and conferencing, the Internet’s high degree of concentration in a small number of Internet Exchange points. Most Internet traffic moves through a dozen or so switching points that aren’t conveniently located to most American neighborhoods; they’re located in “NFL cities.” Internet Exchanges, colocation centers, and backbone Points of Presence (POPs) are the key resources that "connect people to the Internet,” and there aren’t nearly enough of them. As the Internet is a mesh of diverse networks rather than a black box, getting to a place that provides connections to multiple networks is its most significant bottleneck. 

Which future is more likely, continued competition between the telco and cable companies, or complacent cave-in by one side to the other? My bet would be on the former, but there’s no reason to take either scenario as a fait accompli. Nobody can actually predict the future, so the prudent course is to do exactly what the National Broadband Plan recommends: track and compare the evolution of pricing and service. That’s not nearly as exciting as working ourselves into a hysterical lather by reducing complex issues to simple black-and-white terms, manipulating the public, and organizing mobs to go torch the cable companies, but it’s actually quite strongly in the public interest.