A different perspective from Phil’s recent network neutrality posts.
First, I’ll agree that network neutrality is a poor second best to the separation of content and conduit we used to have under the old Computer Proceedings that allowed thousands of independent ISPs to bloom and provide competition and independent service.
That said, there is a reality here. Phil is right this is a complicated issue. But here’s why I think we need network neutrality, defined as a prohibition on letting the handfull of vertically integrated companies that control high-speed access leverage take third party payments for “premium” delivery.
First, a few terms. Phil is right again that the internet isn’t “neutral” in the sense of treating all packets the same. But, as market analyst Blair Levin observed in testimony on net neutrality back in May, the greatest danger to optimizing value occurs when you have a non-transitory bottleneck that restricts economic growth. Here, any individual subscriber faces a relatively small number of providers of high-speed broadband. In some markets, that number is limited to one. In most markets, at least for residential services, it’s two (telco dsl and cable). Some markets may climb higher. None reach the four equal-sized firms that the standard antitrust metric (“HHI”) consider “moderately concentrated.”
Now add in that the largest residential providers are vertically integrated (they provide services that compete with what others want to provide over the network). Worse, they are dominant in a relavant neighboring product market, creating opportunities to leverage this dominance and incentive to prevent competition to their dominant product market emerging.
To translate that into English, cable companies have incentive to prevent real video competition from emerging via the internet, and telcos have incentive to prevent real voice competition. Both companies have the capability and incentive to protect their core video and voice markets, as well as the usual incentive to maximize profit by extorting money from both subcribers and “upstream” content providers. For anyone who has followed the history of the cable industry, this is very well-trodden ground.
To get around the fact that we have different ways to push bits faster, I distinguish between “subscriber tiering”, where I charge customers more for a bigger pipe, “producer provisioning” in which third party providers invest in methods to deliver stuff faster (like peer-2-peer). I call a high-speed ISP charging third parties for “premium” delivery to its customers “Whitacre” tiering, in honor of the Chair of AT&T who popularized the idea.
I think a world that permits subscriber tiering and producer provisioning provides a virtuous cycle of ever-expanding bandwidth and innovation in delivery systems. By contrast, I think permitting Whitacre tiering locks us into a pathetically slow architecture. You can read my more complete analysis here. Briefly, and a little technically, allowing ISPs to charge third parties for “premium” delivery reduces down to an auction for speed among numerous providers indifferent to actual user desires. Whitacre tiering therefore creates an incentive for providers to restrict bandwidth to the point where they maximize the value of the “speed lane.” Worse, the ability of the ISP to slow traffic removes the incentive for providers to innovate or pay others(like Akami) for faster transport. Why bother, if the ISP can neutralize the advanatge and capture that revenue for itself? Finally, this will add a whole new level of expensive transactions for third party providers.
Meanwhile, I am also worried that the ability to charge top dollar for a delivery advantage has very real consequences for democracy. My lengthier analysis here. Again, I’m operating on the assumption that allowing ISPs to offer a premium service amounts to an auction for speed, resulting in much higher prices for transmission of high-bandwidth content with no ability of the user to control preferences. I am not even assuming a deliberate intent to favor a particular message, which cable operators already do in the cable advertising world. I’m basing this analysis on broadcaster behavior under the current rules that require them to offer the same terms to every candidate.
Finally, I am concerned about something I call virtual redlining. I assume that third party providers will not spend money trying to reach customers they don’t find “desirable”, and that access providers will accomodate this. For analogy, consider how advertising rates in broadcasting are now linked to demographic ratings analysis. Companies pay top dollar for advertising on television shows that attract white 18-35 year old males, and pitch their products for that market on shows that command that demographic. Other shows with different demographics get different advertisements and advertisers pay less for these “less desirable” customers.
I see no reason we will not see the same phenomenom if third parties must pay for premium access to customers. Third party providers of content and services will seek to maximize delivery to “desirable” customers in the wealthiest zip codes. Providers and services targetting particular demographics will pay greater or lesser premiums based the percieved desirability of maximizing access to those zip codes. I leave it as an exercise to the reader to determine if this is a positive market efficiency or not.
I’ve read Phil’s “Third Way” paper and agree with much of what he says. My problem is that Phil seems to think that bad results would occur only if ISPs acted from an anti-competitive bias. Accordingly, increasing transparency (for both efficiency and ease of enforcement) and prohibiting anticompetitive discrimination solve the possible problems while allowing markets to capture efficiencies.
I don’t think that’s true. Bad results can happen, as I hope my arguments above (and spelled out in more detail on the posts to which I have linked) show, even if we follow Phil’s recommendations of having clear and neutrally available “premium” policies. Because allowing ISPs to sell premium access to thrid parties, indifferent to user preferences, creates its own set of negative consequences independent of any discriminatory intent.
I am also deeply suspicious of the supposed superiority of post hoc enforcement to prophylactic regulation. (Unsurprisingly, I’ve written at length on this subject as well.) Prophylactic rules let parties know in advance what they can and can’t do. Combined with swift enforcement for clear violations, they save a lot of time in negotiations. By contrast, making rules on a case-by-case basis via after the fact adjudications encourages parties to keep pushing the envelope on prohibitted behavior until they get smacked by the agency. Then they complain that the agency didn’t give them fair warning.
Please note that prohibiting “Whitacre tiering” does not mean the end of innovation or even the end of tiered provisioning. What I object to is allowing third parties to establish user preferences. I, at least, would be happy with a system that lets users designate certain third parties for premium delivery (essentially a specialized form of “subscriber tiering”). For example, this could be something like Comcast’s “power boost,” but where a user pays to get available excess capacity for downloads on a transaction by transaction basis. This would concievably capture many of the advantages that Phil and others forsee in differentiated offerings, but not create the negative incentives and negative consequences I’ve described above. It would also resolve the problem of the so-called “power users,” where a handful of users absorb the vast amount of available capacity on the network.
I recognize that my policy prescription may cut off some potential benefits. Not all deals between ISPs and content or service providers would necessarily be bad. But, trying to predict the behavior of the various actors, I think we have a much better internet, and therefore a much better public policy, if we keep the internet open and don’t let last mile providers get in the way.
Unfortunately, you don't
Unfortunately, you don’t understand the underlying technical issues, even though your analysis is somewhat better than that of your PK colleagues. Here’s why.
You say: “First, Whitacre tiering removes the incentive to build a bigger pipe for the end-user customer. Worse, it creates an incentive not to build a bigger pipe. Why? Because the value of the “premium” service offered to third parties depends upon it being both scarce and necessary.”
Bzzzztttt!!!
You’ve gone and detached the economic value of premium service from the infrastructure that’s necessary to provide it. Premium network services are of two kinds, those that have higher caps on bandwidth, and those that have lower caps on packet jitter. In both cases, the attractiveness of the premium service isn’t a function of its performance relative to some other service, but rather from its performance in an absolute sense. To handle a phone call, I need to minimize jitter such that it’s within the limits set by the human ear and brain, not according to some relative market performance. And when I download a huge HDTV video file, I need for it to download within my attention span, not just faster than I can download it from some other service.
The ability of the network provider to sell premium services is enhanced by increasing the power of the network’s internal infrastructure, not degraded by it. The fatter the pipes inside the Cable Co. network, the more VoIP calls it can handle across its entire user base, and the more fast downloads it can support.
So there is no reason that the service you call “Whitacre Tiering” (obviously to tug at the heartstrings of the jury) would lead to slower growth in network capacity than we’d have otherwise. In fact, it provides Whitacre with an incentive to fatten his pipes because it enables him to get paid for the fattening, and puts the costs on those who stand to gain the most from fat pipes.
It really is necessary to know what you’re talking about when you advocate public policy, Mr. Feld. Otherwise you cause some of them “unintended consequences” that your mother warned you about, and that’s not good.
Generally, I don't answer
Generally, I don’t answer trolls, particularly when they give all the indicia of being paid shills. But this red herring gets raised so frequently that a few comments are in order.
1) Yeah yeah, different services need to be treated differently. this is not news. Latency tolerable in email is not tolerable in VOIP, yadda yadda yadda.
Usually, it is only deliberate distortion or abysmal ignorance that treats this as a trump card. When employed by trolls engaged in deliberate distortion, it is usually said with great fanfare and with snide asides about the purported failure of those — like that non-technical moron Tim Berners Lee or that equally thick-headed Vint Cerf — who just don’t “get” that we need to treat different packets differently.
To list some of the most obvious counter answers: First, no one has proposed outlawing prioritization of packets based on the nature of the service. What gets me worried is — as I said about a gazzilion times above — allowing third parties to set preferences for subscribers. Depsite your feigned disdain for economics, this has some real world consequences that techno-determinism does not solve.
Seoncd, no one has explained why you need to charge third parties in order to achieve this needed prioritization. If ISPs need to charge someone, why not users? I don’t advocate that, mind, but it would more directly address the argument that a few users downloading endless pirated videos are clogging up the system for the rest of us.
2) Anyone who thinks you can divorce technical and economic considerations from one another needs to think again. They drive each other quite nicely. Companies do not invest in the best architectures or the best systems, they invest in the most profitable systems. As a consequence, you need economic incentives to drive companies to invest in better engineering solutions.
And bad news, the economic considerations will always trump in any large, publicly held company. Any engineer who has spent considerable time in a large company has this or that horror story of how a short-sighted economic decision trumped what he or she feels was the better result as a matter of technology.
To review: 1) Your argument does not explain why Whitacre Tiering is either good or necessary, it merely asserts a well-known fact that some packets will require prioritization based on the ability of a service to tolerate latency. This has nothing to do with whether or nor ISPs need to charge third parties; nor do you propose any explanation linking your premise with your conclusion.
2) Whatever my technical expertise, it seems there are some pretty smart tech folks who agree with me rather than you. So unless you assert that Sir TBL is as ignorant as you apparently think I am, I shall rely on him rather than you for technical assesments.
As a final aside, trolls don’t do well in the light of day. And your post displays the classic indicia of trolldom rather than simple dissent. As a group that just outed such a plant recently observed, trolls and other “fake” commenters tend to have well recognized patterns. In your case the intermixed personal insult, the snide aside at supposed lack of credentials, the non-linear nature of the argument, the focus on a few trivial details in the initial post to obscure larger arguments and issues and hopeful drag the respondent (or other respondents) into a debate on trivialities, and, most tellingly the fact that you seem to be haunting pro-net neutrality blogs generally to make similar arguments, are all the indicia of a professional troll rather than a substantive and genuine dissent.
While I suppose I should be flattered that you consider me (and PK of course) worth targeting, I would suggest that it is ultimately a rather futile waste of resources. But far be it from me to stop someone from making a living.
So your response to my
So your response to my criticism of my comment is to jump up and down and call me names, invoke the authority of application programmer Tim Berners-Lee and Google employee Vint Cerf, and to misrepresent the nature of the legislation in question. Let’s try and untangle the mess you’ve got yourself enmeshed in.
1) Nobody’s paying me to comment on net neutrality blogs, to write about it on my own blog, or to give interviews with the media on it. I comment on blogs because that’s what comment sections are for and because I have a particular personal interest in this issue because I’ve been designing networks for 30 years. The twisted-pair version of Ethernet and the QoS enhancements for WiFi are among the systems I’ve helped invent. I use me real name and you can check my credentials quite easily. You, on the other hand, are paid to express the opinions you express, so your “paid shill” claims are classic projection.
2) You claim: “First, no one has proposed outlawing prioritization of packets based on the nature of the service,” yet the Snowe-Dorgan, Markey, and Wyden bills forbid levying fees for such prioritization. This is a key point, and one where the chief authority you cite - Berners-Lee - departs from the manufactured mass movement for regulation of the Internet. Berners-Lee says it’s perfectly legitimate to charge more for a higher grade of service, particularly the service required by VoIP and IPTV. So you either don’t understand what your authority has said or you’re choosing to deliberately misrepresent it. So when I assume you’ll simply ignorant of the technical issues, I’m offering a charitable assessment.
3) You seem to be particularly outraged at the idea that third parties might pay ISPs for improved QoS in sessions with the ISP’s direct customers. If we can agree that QoS has value, as Tim Berners-Lee does, then it’s certainly rational to charge for it as a distinct service. It’s not just a matter of moving bits, after all, it’s a matter of moving streams of bits with consistency. Packet networks react to congestion by increasing delay, but this strategy doesn’t work for delay-sensitive applications. The packet network provider can only move large numbers of streams without delay by purchasing more equipment (routers and circuits) and that purchase is abetted by fees for the particular service that needs that equipment. So it’s a perfectly rational transaction for me to pay my ISP to buy the equipment to handle my voice streams appropriately. You almost admit this yourself.
So if it’s rational for me to pay for enhanced service, why is it irrational for third party to pick up this part of my bill on its own, especially when they have a revenue stream to cover that cost such as advertising? The newspaper that I buy for 25 cents an issue costs much more than that to produce, and most of the cost is born by advertising. So what’s magic about whose pocket the coins come out of?
Now I realize your feelings are hurt because I’ve insulted you by pointing to the thin ice your techno-economic analysis skates upon, so you’ve responded emotionally. That response didn’t make you or your employer look good, so maybe you should calm down before ranting again. Or just don’t reply, the better to keep from digging a deeper hole for yourself.
Excellent post, Harold. I
Excellent post, Harold. I particularly like the bit on redlining. Anyone who doubts this will happen should read Phil Napoli’s book Audience Economics.
I’ve mentioned this to Bennett before, and it clearly hasn’t stuck: the repeated, hateful ad hominem attacks are just poor form. Paid or not, your comments (and attacks on your own, incorrectly-advertised “original” blog) sure could add a lot more to the conversation if you spent more bits telling us what you know and less telling us what we don’t.
The only hateful ad hominem
The only hateful ad hominem (thanks for the definition, that was like real helpful to all of us philosophy majors and stuff) on this thread were penned by the excitable Mr. Feld, Herman, so maybe you should consider redirecting your finger-wagging.
I rather doubt that you’d be interested in a discussion of time and bandwidth management with isochronous protocols and how stochastic network processes relate to constant bit rate streams, packet jitter, rate selection, and transcoding but you never know.
What part are you unclear on?
And BTW, I put up the first political activist blog in the world in 1996, and I figure that entitles me to call my current blog any damn thing I want. No go do your homework and stop being such a pest.
In an effort to improve the
In an effort to improve the dialog on the PK boards, I’ve moved my portion of this catfight to my blog. Care to join me there?