Contents

Introduction. 2

The Lay of the Land: Technical Setup of the Internet 2

Peering and Transit 3

Differences Between the Internet and Telephony 3

Content Providers Also Invest Heavily in Networks 4

The Unfair Costs of Imposing a “Fair Share” Model 6

Conclusion. 7

Endnotes 7

The Information Technology and Innovation Foundation (ITIF; Transparency Register #: 923915716105-08) appreciates this opportunity to comment on the European Commission’s consultation regarding the future of electronic communications and infrastructure. In particular, these comments concern proposals to implement a so-called “fair share” policy that, in short, would mandate that content providers, such as Netflix, pay fees to Internet service providers (ISPs) in return for use of the latter’s bandwidth.[1] 

Introduction

The volume of Internet traffic has increased dramatically. ISP proponents of a “fair share” policy argue that this increase, in addition to the fact that the majority of people already have broadband, causes the costs of providing broadband service to rise faster than revenue. Moreover, the European mandate of universal gigabit coverage will necessitate massive investments on the part of ISPs, while the content providers whose business models rely on Internet access—and who regularly send traffic across ISPs’ networks—“free ride” on the existing pipes.

This argument for a “fair share” policy proposal is based on a significant mischaracterization of the existing Internet ecosystem. Rather than content providers sending data unprompted, it is the ISP’s own customer that is paying to have the data delivered to them. Given this, the “fair” route would be for ISPs worried about too much data use to charge heavy data users more.

The reality is that every actor with a stake in the game plays a financial and business role in moving content to its destination. All organizations that put content on the Internet already pay to do that. Content also makes ISPs’ services more valuable—more content availability is a benefit to their business models, not a deadweight cost. Moreover, there is not a one-to-one correlation between traffic increases and costs borne by ISPs; in fact, in a world of ubiquitous Internet use and skyrocketing traffic, it does not appear that ISPs pay significantly more at all.

Static, one-way fees would undermine the flexibility and resilience of the current Internet architecture and erode many of its key strengths. They would also likely be shouldered by consumers. Rather than correcting an existing imbalance, therefore, implementation of the “fair share” proposal would instead impose new, unnecessary costs and burden the Internet ecosystem.[2]

The Lay of the Land: Technical Setup of the Internet

Despite its apparently cohesive global reach, the Internet is a diverse realm of thousands and thousands of networks that connect to each other through traffic and data exchanges. Because the Internet is so large, such connections are rarely direct. The apparently seamless movement of data from source to destination is frequently actually a series of hops from one network to another.

Since the Internet is global but the networks that constitute it aren’t, traffic frequently has to move across multiple networks in the process of getting from source to destination, and an entire ecosystem has developed to support the evolving role of traffic transport and the processes that enable it.

When a creator uploads content—such as a video—to the Internet, that data is encoded and sent in the form of packets. Routers at networks along the way look at packet headers to determine where the data is headed and direct it until it reaches its destination, at which point the recipient reassembles the packets into recognizable content.[3]

There were originally three major players directly involved in content transmission: the end user who requests the content, content and application providers (CAPs) that produce content and facilitate its delivery to their own subscribers, and Internet Service Providers (ISPs), who offer contracts for Internet access to both end users and CAPs. In sending content on its way, ISPs connect to each other and to CAPs at Internet exchange points (IXPs), which are physical locations where traffic is routed among different participants.[4]

As modern-day Internet usage sees ever more transmission of massive amounts of data over vast distances, content delivery networks (CDNS) have evolved to take some of the strain off long-haul transport. CDNs are groups of distributed servers—essentially networks of computers—that each ‘cache,’ or host, duplicated versions of content to bring it closer to end users and, therefore, require less extensive use of ISPs’ own infrastructure.[5]

Peering and Transit

Network exchanges are a crucial part of moving data from source to destination. As they direct traffic across the globe, each network uses the same protocol—Border Gateway Protocol (BGP)—to map a route across the Internet.[6] This protocol is what networks use to find the shortest, more efficient route in each instance of data transmission.[7] Since the Internet is so vast that no one map can be a comprehensive accounting of the entire thing, neighboring BGP servers often collaborate by sharing their maps with each other, and sometimes they then agree to share traffic.

There are two fundamental models for how networks connect when they need to exchange traffic: peering and transit. Peering takes place when networks exchange traffic directly, frequently because they are exchanging roughly equivalent volumes of traffic. Transit happens when one network agrees to carry another’s data, and help distribute it, for a fee.[8] The scope of an ISP’s network coverage, which can range from limited in reach to global, categorizes ISPs into tiers: They range from Tier-3 providers, the smallest, most local networks, to Tier-1, near-ubiquitous networks that function as the Internet backbone.[9] When payments are involved, they generally flow upward through the tiers.

Though handshake agreements are the norm, ISPs are certainly free to request settlements during any traffic exchange if they believe that is the best business decision for them. In fact, in exchanges among different-tier operators, the larger often takes a fee from the smaller one in exchange for its transport of the latter’s data.[10]

As long as operators are exchanging roughly equal amounts of data, settlement-free peering is generally mutually beneficial. Both parties often choose to exchange data fee-free and only cover their own network and operating costs. This practice is so common that in a 2016 Packet Clearing House survey of over 10,000 networks, over 99.9 percent of available traffic-exchanging agreements were settlement-free “handshake” deals devoid of any fee or formal procedure.[11] 

Content providers have forged their own peering relationships in their quest to get content to as many potential subscribers as possible. As traffic has grown, CAPs also increasingly invest in their own content delivery networks or pay ISPs to host their content as part of their own mandate: to deliver the best experience to consumers.[12]

Differences Between the Internet and Telephony

In the heyday of fixed telephone operators, which were essentially regulated monopolies with interconnection rates set and standardized by the ITU, mandating fees in a particular direction fit more naturally within the existing framework.[13] The “calling party pays” model imposed on telephone operators, under which the calling party was responsible for handling a fee, was more a standardization of the norm than a revision to the existing structure.

But the shape of the modern-day Internet represents a significant departure from the early days of telephony. On the Internet, data travels across multiple diverse nodes and dynamically seeks out the best route as it travels, rather than in the straight, direct line made by a telephone call. This flexibility is a key part of what makes the Internet strong, innovation-friendly and resilient.[14] Because the Internet Protocol is very simple, it can be run over a variety of networks, so changes to the material of our networks (for example, from copper cable to fiber) can take place while the rest of the network runs uninterrupted.[15] This also means that it is highly scalable, a major factor in the Internet’s explosive growth since the 1990s and a quality of growing importance as we rack up a list of critical uses for the Internet and the Internet of Things.[16]

Finally, the relative interchangeability between different interconnection agreements, and the ability for participants to forge relationships and choose new routes as necessary, are valuable for two reasons. First, this setup facilitates near-constant optimization of data transport based on the conditions at that particular moment. Second, such decentralization creates an Internet resilient enough to be able to cope with sudden or unexpected events like changes in demand for Internet use.[17] Without this system of backups, outages or technical issues anywhere along a route might take entire swaths of users or data fully offline, and an Internet reliant on things remaining just so might not withstand a massive upheaval in the way we use the Internet—like the COVID-19 pandemic.

Content Providers Also Invest Heavily in Networks

It is true that broadband providers invest heavily in their own networks, but other actors do the same. Content providers in particular have a keen interest in allowing the Internet to run smoothly, and since their subscribers expect high-quality, speedy access to content, CAPs are highly incentivized to play a role in providing that quality. One way in which they do that is by investing in networks and supplementary platforms themselves.

In fact, a recent Analysys Mason report finds that from 2011-2021, CAPs invested $883 billion in digital infrastructure—including on data hosting and transport and on delivery networks that both support the delivery of the CAP’s content and, incidentally, ISPs’ business in general.[18] Between 2018 and 2021 alone, CAPs increased their spending by 50 percent.[19] Many of these expenditures go towards supporting data carriage and moving it closer to consumers, where otherwise ISPs would shoulder the costs of higher data transport fees. Altogether, investments by CAPs save ISPs over $5 billion annually in transit and network fees.[20]

In particular, CDNs are relatively newer players in the online sphere that emerged to ease the load of growing data. Uniquely equipped to handle exponential traffic increases, CDNs and other intermediaries are one natural solution to greater traffic flows that would otherwise have caused higher network costs and higher latency.[21] In fact, their use has become essentially nonoptional to the content provider that wants to be competitive, and there are currently around 170 CDN companies globally.[22]

End users see direct benefits from the use of CDNs: First, the distance between the user and the newly stored content is smaller, which results in shorter latency, better reliability such as reduced packet loss, and a lower cost of transport when a consumer requests that content.[23] The size of the file can also be reduced and therefore be able to load more quickly. CDNs even provide additional resiliency against the increasing threat of cyberattacks by duplicating content so its availability isn’t contingent on the functioning of a single server.[24]

ISPs benefit from CDN use for many of the same reasons that consumers do. Content stored by CDNs is closer, cheaper, and easier to collect. Each data exchange conducted through a CDN is also one less transaction that needs to be conducted through transit, which some ISPs would otherwise have to pay for.

In addition to the direct benefits experienced by ISPs and the improved consumer experience, the widespread use of CDN platforms ultimately benefits the rest of the Internet ecosystem as well. Their use frees up capacity on other means of transport for content that isn’t run through a CDN.[25]

Because CDNs are so crucial to the functioning of the modern-day Internet, content providers whose business models hinge on successful data transport often play a direct role in their implementation. Content providers frequently pay CDNs to host their content closer to the end users, which improves quality for their customers and expands the CAP’s reach past a regionalized area. Sometimes they even invest in their own, which benefits them—because the CDN can then be designed specifically to best accommodate their own content—and reduces the cost of data transport for ISPs.[26] Netflix, for example, has its own CDN called Open Connect.[27]

Altogether, an entire ecosystem has converged around CDNs. There are preexisting CDN market players whose only business objective is to deliver content. But as growing traffic intensifies the need for regionally delivered, duplicate data, CDNs are also increasingly owned and developed by ISPs and CAPs alike.[28] Content providers also play an active role in transporting their own data through extensive peering relationships with ISPs and by connecting at IXPs where necessary.[29]

On the other end of the spectrum, submarine cables—a fundamental enabler of overseas data transport—are a holdover of the traditional telecommunications days that remains critical to a healthy Internet today. Even as much of the process of data transport begins to center around the last mile and regionalized areas around consumers, the importance of the Internet backbone remains, especially in traversing oceans to connect continents. Here, too, content providers play an increasing role: Many of the biggest content providers are owners or part owners of multiple submarine cables.[30] Amazon and Microsoft, for example, have each invested in five or more submarine cables; Meta and Google, each over a dozen.

Moreover, network-related costs for ISPs have not increased over time alongside the increase in traffic volumes, largely because many network-related costs are unrelated to the volume of data traffic. To the extent an uptick in data would have naturally increased ISPs’ costs, this has been counterbalanced by savings through economies of scale, technological innovation, and investment by both ISPs and CAPs in digital infrastructure. It is further counterbalanced by increased investment in and transition to fiber infrastructure, which is essentially traffic-proof.[31] In fact, the amount of total traffic nearly tripled from 2018-2021, but telecom operator expenses remained largely stable.[32]

In sum, though ISPs make large and important infrastructure investments that enable today’s Internet to function, those investments are not the exclusive input to traffic delivery. Other players contribute to ongoing network costs, and ISPs themselves do not face proportional additional costs due to traffic volume increases.

The Unfair Costs of Imposing a “Fair Share” Model

Taken together, all of this means that imposing a “fair share” model would be an unnecessary solution to overblown problems based on misleading narratives about the setup of today’s Internet. In fact, beyond addressing an issue that does not exist, imposing these fees would actively harm the existing Internet ecosystem and create new problems.

As described, Internet traffic delivery relies on a constellation of investments and negotiations which are the business decisions of ISPs, CAPs, and end users alike. As in any complex market, the prices that emerge from free negotiations are more likely to coordinate scarce resources than centrally mandated ones. Each player in the market has its own business goals, and price signals help them manage resources wisely to achieve their ends. CAPs must ensure their content will get to their consumers just as ISPs must construct networks to which consumers are willing to subscribe. The two industries need each other, and neither is a hapless victim of the other. An ISP is free to negotiate advantageous peering and transit arrangements, but its failure to get the price it wants from those negotiations is not a reason for the European Commission to mandate a regulated rate. Not getting the deal you want doesn’t mean it isn’t fair.

But beyond the inherent fairness of freely negotiated contracts, “fair share” policies would have far-reaching and harmful consequences for consumers.

From a purely practical standpoint, there is a preexisting financial relationship between content providers and consumers (in the form of, for example, a household’s Netflix or Disney+ subscription). Artificial fees imposed on CAPs would raise their cost of doing business which would, in turn, drive up consumer prices.[33] It may be reasonable to argue that prices should increase to fund infrastructure, but it is folly to hypothesize an increased cost that affects only the producers of content and not their customers. Fees imposed on tech companies pursuant to a “fair share” policy would ultimately be a tax on consumers.

Perversely, Analysys Mason finds that network usage fees would also ultimately raise costs for many ISPs and therefore their customers as well.[34] Because CAPs currently pour such a large amount of their own funds into network maintenance, network usage fees looking to extract more donations would simply disincentivize them from contributing as they have been doing. This would effectively raise the cost of doing business for ISPs, which would have the further effect of reducing competition by pushing smaller ISPs out of the market altogether. Anticipating this, groups including Europe’s MVNOs and Euro-IX, a global community of IXPs, are already collectively opposing the “fair share” policy.[35]

One-way mandatory fees like those the “fair share” policy looks to impose would also misrepresent the direction of value flows on the Internet. Put simply, people subscribe to Internet service to be able to access content. This is in stark contrast to old-school telephony where the direction of value flows was obvious (the person who initiated the call generally did so unprompted). Instead, content on the Internet is generally sent out to an end user at the user’s request—such as when somebody attempts to watch a movie on Netflix. If a few large content companies command the majority of traffic, this reflects the value society actually places on them. In many cases, these companies are likely a big part of the draw of subscribing to the Internet in the first place. A “fair share” payment scheme would effectively be CAPs compensating ISPs for doing their part in keeping the latter’s customers happy and engaged which is, in turn, the basis of the ISP’s value proposition: the ISP’s “pipes” are made valuable by what is on the other end.

To be clear, no part of the above argument suggests that CAPs should get a mandatory free ride on ISP infrastructure. It does indicate that the parties should work out how to assign costs among themselves based on economic realities rather than running to regulators to assign them based on political expediency. There may be good reasons to think that, in the abstract, CAPs do not always have the right incentives to send traffic in the most efficient way. In such cases, any deal in which the ISP charges a fee to force the CAP to internalize the costs of its behavior would likely benefit all users of the network. Enabling contracts and changes in price to accurately reflect the relative scarcity of resources in the Internet ecosystem would be a positive regulatory step. But the fact that the parties don’t strike such a contract reflects the underlying value propositions of each, and that is not an inherent policy failure. The EC should not impose by regulation what an ISP failed to achieve by negotiation.

Conclusion

The Internet ecosystem has been fine-tuned by time and trial-and-error and grown into the behemoth it is today, which has in large part been made possible through the potential for dynamic bargaining between participants. All players have an incentive to invest in network infrastructure and play a role in ensuring the movement of quality content. Traffic-based mandatory fees would introduce a perverse incentive wherein popular content providers would be punished for providing outsized value to the Internet, the largest ISPs would get a regulatory thumb on the scale, and consumers would bear the brunt of the costs.

CAPs should and do invest in the infrastructure their business models rely on. ISPs do not, however, deserve an extraneous fee for their participation in a mutually reliant ecosystem that benefits all parties. To mandate one would be to blithely ignore the strengths of the Internet’s current model in favor of preferencing certain players at everyone else’s expense.

Thank you for your consideration.

Endnotes