Last week’s squib on net neutrality provoked replies from Ezra and my two favorite Caucasian Lees: Tom and Tim (no relation).

First, I want to concur strongly with Tim’s general point: As Robert Frost reminded us, all metaphor breaks down somewhere. (Otherwise, one supposes, we’d call it “literal description.”) Certainly the history of common carrier regulation can and ought to be used to illuminate contemporary cases, but it does sometimes feel as though the debate over net neutrality is devolving into a war of analogies—my own post being a case in point. The Internet really is sui generis, and it’s probably more useful to discuss it, and having to wrangle over which aspects of one medium or form of “discrimination” between customers it’s necessary to capture in a comparison to another medium. This adds another layer of confusion, such that a number of commenters, here and at Ezra’s site, read the choice of a loose analogy stressing one narrow point as betraying a failure to understand the mechanics of “policy routing”—which I think makes this the first time I’ve been accused of not being enough of a computer geek.

That said, before we abandon analogy entirely, let me respond to this at Ezra’s:

What market interest is being served by the increased accessibility of Pizza Hut? After all, we want our pizza places competing on a variety of metrics, from deliciousness to delivery speed to courtesy. How much bandwidth they can purchase, however, is not one of them. Indeed, a preventable situation in which subpar pies are propagating because Luigi is purchasing more tube space than Mario is…a bad situation.

Forget for a moment whether this has any applicability to net neutrality: this is just wrong. There’s no reason competition along the dimension of delivery speed is good, but competition on the dimension of customer access produces a “bad situation.” Some places specialize in producing high volumes of mediocre but quick or cheap pizza, others in pricey gourmet stuff you’ll have to wait for. Economies of scale may mean larger places have an advantage on some of these dimensions and a disadvantage along others. But there’s not some standard independent of the preference weightings revealed b customers’ choices by which it makes sense to say this or that balance of models is “wrong.” If a place fails to buy enough phone lines, or hire enough people to man them, or to attend to customers in the physical store, people may go elsewhere. That’s what either sends a market signal to firms to build up that end of their service.

Now, to move back to the actual issue at hand: I think the reason people see this as different is that it sounds zero-sum. There’s an obvious consumer benefit when businesses are motivated to pony up for more phone lines or fatter pipes to better serve their customers. But the people fretting about various non-neutrality scenarios typically talk about firms paying to give their packets preference over their competitors’, which sounds more like “competition” to deflate the tires on the other pizza joint’s delivery cars. And that’s what makes it seem plausible for Cory Doctorow to use phrases like “protection money” in describing the fees companies might pay ISPs: It’s a payment to be spared an unnecessary slowdown.

Now, I’m not saying you couldn’t have a destructive arms race like this, but I do think there’s been a tendency to slice the question in the following intuitive but probably mistaken way: Paying for bigger or faster pipes, or competition along these lines, is fine; paying to get the fastest available access to an ISP’s customers while the ISP deliberately slows down most other traffic is like tire-deflating. And I think that’s not necessarily the relevant distinction. (Tom’s post thoughtfully distinguishes between beneficial and harmful or rent-seeking types of policy routing without, I think, falling into this trap.)

Here’s why. Suppose an ISP wants to build out infrastructure to support 100mbps service in an area that doesn’t currently have it. Problem: There are customers who might like that higher speed access for a few purposes, like streaming movies, but not enough who are prepared to pay the premium to upgrade their whole connection, so it’s not cost effective for the ISP to make the investment. One solution might be metering: You let customers pay for the bandwidth they use, paying a bit more for bursts of higher speed needed to access specific sites with a lower flat rate for the majority of the time, when they’re just reading news and checking email. The problem is that consumers seem to have largely rejected metering: People want to pay one rate for their access, and not have to think about their usage level on a day-to-day basis.

The other solution is for the sites the customers are trying to reach, the ones that many people will only find worth using if they connect at very high speeds, to subsidize the infrastructure buildup, then recoup the layout in the form of increased revenue from subscription fees or ads or however else sites convince investors they’ll make a buck. Of course, that only makes sense if the infrastructure buildup doesn’t just result in cheaper and faster across-the-board access for the customers of the ISP in question; the benefits have to be internalized by the company making the payment. Now, in terms of how this was originally framed, once the infrastructure is there, that means “slowing down” any traffic not coming from the favored (paying) companies to lower speeds than the new and improved pipes can accommodate. But in the case I’ve described, of course, that’s not zero sum because the fatter pipes are only there due to the cross-subsidy from the favored companies. This would be more intuitively clear if, say, the companies had subsidized the construction of a physically distinct set of fatter pipes that only carried their traffic—except that’s sort of insane and inefficient, and nobody’s actually going to build a network that way.

This is a just-so story, of course: I’ve described a way that preferential access involving “slowing down” some packets and passing others along at full speed might be optimal. That doesn’t show it’s the most likely use of policy routing. On the other hand, some folks seem incapable of imagining any beneficial use, so it may be worth at least gesturing in the direction of one. And it matters because if there were only downsides to a discriminatory net, then even if the horror-hypotheticals or actual abuses were highly implausible in the first case or highly uncommon in the latter, there would be no harm in regulating. If there are pros and cons, it won’t do to just point out what kinds of bad things might happen. You need to figure out how serious and common they’re going to be relative to the beneficial cases.

So the real question, as Tom emphasizes, is not whether there are beneficial uses of net “discrimination,” but which type will predominate. That, I think, turns largely on the extent to which local broadband incumbents are subject to the discipline of genuine competitive threats within regions, whether particular ISPs can offer up enough people to enable them to extract rents, and so on. There’s certainly a range of defensible views there. But two things I think need to be borne in mind to prevent hyperventilation: (1) Some of the worst case scenarios are precluded just by ordinary truth-in-advertising rules: You can’t sell people 10mbps access and then routinely and deliberately make the majority of sites slower than that baseline; (2) The desire to extract rents is not the same time as the market ability to extract rents. So for the most part, the worst-case-scenarios are beside the point. Monopolists in any market will extract rents when they can, and firms in a competitive market will daydream about extracting rents but not typically have much leeway to do so. The interesting question is not about the various and creepy forms rent extraction might take; it’s whether there’s enough looming competition in broadband markets to check those attempts, whatever form they take.