Economist/NYT opinion writer Paul Krugman coined the term “Zombie idea” to describe an idea that, despite being repeatedly refuted with evidence, keeps coming back. Not surprisingly, zombie ideas typically have powerful constituencies that benefit from the adoption of the zombie idea they push, and who invest a great deal of money and energy continuing to resurrect the idea each time it gets killed.
This Halloween brings us the return of the “content companies should pay money to last mile networks because they use more resources” idea. I first explained why this was a dumb idea back in 2006, after then-AT&T CEO Ed Whitacre explained that he wasn’t going to let companies with content his subscribers actually wanted “use his pipes for free,” Despite lots of reasons why this is a Dumb Idea that would end up seriously reorganizing the internet economy for the worse, you find carriers around the world reviving it because $$$. This is particularly true outside the U.S., where the argument also gets caught up in debates about American dominance of internet content and popular culture. There is a separate U.S. flavor, supported by folks like Brendan Carr, about charging “big tech” to build out broadband infrastructure, which I’ve also previously criticized. But the non-US flavor has been gaining traction as a function of the “Techlash” and therefore needs some in depth discussion — especially since we can actually see the predicted bad consequences play out in real time in South Korea.
Back in 2016, South Korea adopted a new interconnection rule based on a long-standing telco compensation rule called “sending party network pays” (SPNP). As I’ll explain in detail below, SPNP has deep roots in the whacky world of telecom “settlement” (the fancy word for who pays whom in international calling) how networks compensated each other for exchanging traffic. Those opposed to adopting this approach predicted (based on about 100 years of history) that it would prove impossible to enforce without super intrusive government oversight and would introduce severe latency into S. Korea’s networks as the “sending networks” (such as Netflix, but also gaming companies and others with high resolution visual content) routed traffic in clever ways to avoid paying significant charges. To the surprise of no one except the advocates for the proposal, the predicted badness happened. The cost of transit skyrocketed, latency dramatically increased, and the Korean government keeps needing to consider new and more intrusive ways to (a) stop companies from avoiding the fees to ISPs while (b) trying to target foreign content providers while protecting domestic uses they like — such as video chat and video games.
Despite this real world example, and an impressive array of folks explaining in detail why they totally hate this stupid idea, important folks in the European Council (egged on by the EU telcos) continue to think this is a Totally Awesome Idea. So I will explain how we got here, what traditional “sending network party pays” actually means, why this ain’t it, but even setting that aside, why what happened in S. Korea shows this is a Really Dumb Idea.
More below . . . .
So This Comes With Some Super Long Background About How Networks Exchange Traffic So We Can Understand What’s Going On and Why It Sucks, Right?
Absolutely! Figure you’re probably due for 1K words or so of history here to understand what South Korea actually did and why it has worked so poorly. This is Wetmachine, not TikTok.
Please Keep to under 1K words?
I’ll try, but no promises.
Sigh . . . Go Ahead.
To understand what is going on here, you need to understand that building and running networks — especially communications networks — does not work like normal economics. Of relevance here (and discussed in much more detail here), the cost and return on investment for delivering the exact same product (data, voice, or dits and dots of a telegraph) varies enormously depending on things like local population density and terrain. The less dense the population, the more expensive to deliver voice, and the lower rate of return. This makes funding build out to rural areas challenging. Telecom networks also have a long history of being government monopolies in most of the world (basically starting as a subdivision of the postal service). Finally, the phone network historically operated as a legal monopoly or as a de facto monopoly (e.g., “Ma Bell“).
What this meant was that telephone companies charged customers for building out networks and providing service, but also came up with other complicated ways to charge each other to actually complete phone calls. Because of the magic of network economics, these charges had virtually nothing to do with actual market value or cost and much more to do with things like monopoly rents when not regulated and government policy when regulated — and sometimes both. One of the boondoggles on this was the idea of “sending party network pays.” The theory being that if a foreign network tried to reach a customer on my network, I incurred a cost (holding open a circuit and line) — despite the fact that I actually built the freaking network to perform this function.
Of course, when the market power and/or political power ran the other way, you got “receiving network party pays.” Here the theory is that I am getting a benefit from the other network deigning to deliver its traffic to me, so I should compensate the originating/delivering network for the privilege of receiving a call from the other network — despite the fact that the subscribers of the other network presumably subscribe so they can make these phone calls. None of these really has anything to do with economics, of course. Some of us are old enough to remember back in ye olde days when cell phones were new and virtually all the traffic was going from wireline networks to cell phone networks and the cell phone networks needed to interconnect with the wireline networks or they did not have a business (since without interconnecting to the wireline network, you could basically talk to all 5 other people who had a mobile phone). Remember how you had to pay both when you made the call and when you received the call (or text)? That was because the wireline networks had both the political and economic power, so they were able to get both “sending party network pays” and “receiving party network pays,” despite the fact these are completely mutually exclusive theories of how to apportion the cost of the network (which, I will point out again, is also being paid for by the subscribers).
Mind you, it wasn’t all extortion all the time. Government oversight, when it worked, was about making sure that the rates were “just and reasonable” (i.e., they actually bore some relationship to the cost incurred for completing the call) and as a means of subsidizing government policy without people noticing. For example, in the U.S. we forced urban networks to pay high access charges and termination fees to rural networks (because, as noted above, those networks cost more to run for lower return). This system of hidden subsidies helped to keep telephone service affordable for rural America. While that sounds like a tangent, it actually gets more relevant as we move along so keep it in mind.
Then comes packet-switching and the Internet (which actually was capitalized back then, as the singular, global “network of networks” and not just “that thing I use to tell people they suck because of their politics/choice of sports team/choice of movie or streaming service”) which — for reasons I will actually spare you — completely upended the economics of all of this. In exchange for being just “best efforts” and fully outside the whole telecom/government monopoly thing, networks developed lots of ways to move information around. (If you are wonk enough to be curious, I recommend this here, this here and this here.) Due to a peculiar combination of actual government regulation, threatened government regulation, and funky market economics, the internet (now reduced to a common noun) junked the whole 100 years of complicated telecom pricing and has largely worked out for moving bits between networks through a combination of individual contracts and “bill and keep” or “free peering” (two names for a system where the networks interconnecting and exchanging traffic decide it is too much hassle trying to keep track of who sends whom more traffic or which derives more benefit or has more power, so they just exchange the traffic with each other for free). We also have “content delivery networks” (CDNs) which take high volume content, host it close to the big last-mile network, and deliver it when requested to cut down on latency. For reasons largely having to do with history, CDNs generally pay networks to interconnect. (You may remember a big fight between Netflix and Comcast and other big last mile ISPs a decade ago? That was all about transit v. CDNs and who got to charge for what when.)
Anywhoo, the point of all this background is that for over a century (and to some degree continuing to today) telephone networks had a funky way of charging each other, generally overseen by governments to some degree, and used to profit telephone networks with market power and political power while simultaneously supporting various types of government policy — such as building out networks in high cost rural areas. Then along came the Internet and most of that got trashed. Despite non-stop efforts by carriers for about 15+ years to try to get payments from content providers as well as subscriber fees from subscribers, the internet mostly avoided this unpleasant fate. Until South Korea in 2016.
Are We Done With the Historic Background?
Yes, and I kept it to 1K words. Happy?
So What is Going on in Korea?
You can see a good basic summary (along with an initial impact statement) by the Internet Society here. See also Open Net Korea here. Prior to 2016, S. Korea’s internet worked like most of the OECD countries. ISPs and CDNs negotiated with each other for payment, and primarily relied on free peering among the 3 dominant ISPs. But in 2016, S. Korea became the first OECD country to impose traditional “sending party network pays” compensation rules. Importantly, these are mandatory rules. ISPs cannot decide to simply waive the fees and continue to provide “bill and keep” interconnection. Why the S. Korean government — through it’s telecom regulator — decided to impose this is unclear to me, as I cannot claim any expertise in S. Korean politics. From what I hear, however, it has a lot to do with the fact that the largest content producers (or “content application providers” aka CAPs) are U.S.-based companies like Netflix and Google. The idea appears to have been that rather than pass these fees on to S. Korean subscribers, the ISPs would turn around and bill the CAPs. As is generally the case in such schemes, there is no actual obligation to pass the savings from these revenues on to subscribers — so it’s not like broadband prices in S. Korea went down.
Critics warned that the result would be CAPs moving their content out of Korea and breaking it up among international carriers to reduce the amount of traffic exchange to any one ISP and thus avoiding (or at least minimizing) the access charges to the S. Korea end user networks. This is bad because it increases latency. Rather than store and transmit data in the most efficient way possible, which might mean transferring lots of data from one ISP to another (and thus incurring a huge SPNP fee), CAPs began storing and routing data to avoid the fees. This meant doing things like requiring S. Korean ISPs to each interconnect directly outside of S. Korea and thus avoiding the S. Korean SPNP rule (since once the data was subject to S. Korean jurisdiction in S. Korea, it was already “inside” the ISP and therefore not ‘exchanged’ and subject to SPNP). This means the packets requested take longer to arrive, and may go through multiple “hops” to reach their ultimate destination. These all increase latency. The increase in latency, in turn, messes with applications that require realtime responses (like voice or video games) and causes jitter or buffering in high-bandwidth content like video.
If this sounds familiar, it’s because we had a similar issue here in the United States over rural call completion as carriers used voice over IP (VOIP) routing to avoid the higher termination fees meant to subsidize rural phone systems. This is what made it so easy to predict why bad stuff would happen when S. Korea decided to impose telco SPNP on broadband. (that, and a basic understanding of economics and the technology.) Companies, as rational actors, try to minimize their costs. Unlike ye olde circuit switched phone network, which offered limited circuit switched pathways to complete calls (although we had fun games and shenanigans there as well), IP routing offers lots of flexibility in delivery. This usually results in companies trying to avoid latency by finding the most efficient way to deliver packets. But if introducing latency by routing tricks saves oodles of money, then rational actor profit maximizing firms gonna rationally act to profit maximize. Result, according to this OECD Report at page 51, is that S. Korea now has the highest latency of all OECD countries, and the problem is getting worse every year (in contrast to significant drops in other countries).
So rather than get a big payoff, S. Korean ISPs had to incur new expenses (both to build out to the new, offshore traffic exchanges and to set up the accounting system for SPNP) and degraded network performance.
So S. Korea Realized This Was A Bad Idea and Went Back to Peering/Transit with a side of CDNs?
So S. Korea Doubled Down and Tried to “Fix the Problem” by Getting Even More Heavy Handed?
Bingo! Unsurprisingly, as S. Korea doubled down to address “regulatory arbitrage,” the solution looked more and more like “lets stick it to foreign content providers while trying to protect our own services” and less and less like a traditional “sender pays” telecom pricing network. In 2020 S. Korea passed a law amending the Telecommunications Business Act (TBA) to require “value added telecom service providers” (VSPs) (which I think are basically content application net6works like Neflix, along with other data providers) to directly pay ISPs and give up on pretending this is about traffic exchange. The 2020 law requires content providers that account for 1% or more of S. Korea’s average daily usage and have at least 1 million subscribers to take measures to ensure that their service is stable within the country. Apparently, this didn’t address the problems (or result in a big enough payoff to the S. Korean ISPs), so the S. Korean parliament is now considering even more bills to basically force content providers to enter into contracts with ISPs that “pay their fair share” of supposed expenses generated by ISP customers wanting access to their content.
As far as I can tell (given I don’t read Korean or know Korean law), and as is usually the case in these schemes proposed by carriers, nothing in the law or proposed bills actually requires any sort of cost-based accounting or any requirement to either pass the money back to S. Korean ISP customers (who were theoretically paying the “unfair share” of the ISP infrastructure investment via subscription fees) or otherwise invest the money in deploying new networks or something. It’s ye olde AT&T’s Ed Whitacre “you ain’t gonna use my pipes for free.” Basically, the more S. Korea tries to tork itself in knots to get Netflix and other content providers to pay money to S. Korean ISPs for no good reason beyond “we wants it, Precious!” the more it mucks up the entire economy of the internet in S. Korea. If you want a more detailed analysis of the many different ways S. Korea’s effort to create toll roads for content on the internet replaced the previous “virtuous cycle” with a vicious downward spiral, you can see this ISOC analysis here.
Dude, You Are Being All Free Market Today. You Sure This Is Wetmachine? And Didn’t Something Like This Work To Solve Rural Call Completion?
As I keep repeating over and over again, it’s not like “regulation bad” or “regulation good.” Life is trade offs, and policy (whether regulation, deregulation, subsidies, or what-have-you) is about aligning incentives to maximize the likelihood of good results and minimize the likelihood of bad results (what constitutes ‘good’ or ‘bad’ depends on your desired policy outcomes). It’s also a question of how much you are willing to pay to meet your goals. You need to understand markets, not worship them.
We could solve (to some degree) the rural call completion problem with a bunch of regulations — over the course of about a decade — because it (a) addressed an actual real problem (people in rural areas needing a working phone system) rather than just a cash grab, which made it worth it to put in the time and effort and find a sustainable consistent solution; (b) even with voice-over-IP instead of circuit switching, voice service is still much more straightforward, and more regulated, than broadband data routing; and, (c) we were actually trying to fix a systemic problem and apply the same rules to everyone, rather than soak a handful of popular (and foreign) services for the benefit of a specific industry segment.
None of those factors apply in the S. Korea case (or the EU). Prior to 2016, we didn’t have an actual problem. Then ISPs got S. Korea to use the power of its government to try to get a windfall from foreign content providers. Initially, ISPs dressed this up in traditional telecom rhetoric, but eventually they dropped even this pretense. As a consequence, it keeps backfiring badly.
Which brings me to one final point that I want to stress, because I know this will get lost in discussion when the policy fight gets back to the U.S. and I (and other Title II supporters) will be misquoted and deliberately misconstrued all over the place. Broadband access (and, I would argue, transit) are telecommunications services, but they are not telephone services. Just like the telephone service wasn’t a telegraph service. This is why I (and others) spend so much goddamn time talking about basic principles of common carriage why even if the basic principles remain the same, the application of these principles to new technologies will always be different and must reflect the real differences between the technologies to protect the underlying important principles of what former FCC Chair Tom Wheeler called the Network Compact.
We’re at That Magic 3,000+ Words Where You Usually Cut Off. Any Last Thoughts?
Basically, as I predicted over 15 years ago, allowing last-mile ISPs to charge access fees to content providers to reach the ISP’s customers (who, after all, are subscribing to the ISP to get the content) isn’t just a cash grab that leverages the power of the network over the customers (y’know, that whole “two-sided market” thing I spent a chapter describing in my Case for the Digital Platform Act). To quote myself from 2006: “[it] produces really, really awful results from an economic perspective. It gives actors all the wrong incentives, adds new layers of uncertainty and inefficiency to the market generally, and discourages investment in bandwidth capacity at every stage of the network.” Turned out this was totally right. Go me! I am so smart! S-M-R-T! Also turns out that when you have governments impose these access charges rather than simply let ISPs exercise market power, the results are equally bad.
S. Korea is not a model — either for the EU or the US or anywhere else. It is a warning. The fact that this powerful, real world example cannot kill the “sending party network pays/content provider pays/bandwidth hog/whatever you call it just give me the other guy’s money on top of my subscription fees” zombie is a true tribute to the power of carrier lobbyists and the magical thinking of certain policymakers.
Stay tuned . . . .