The internet infrastructure’s turn to content

One of the internet’s key features is that it keeps networking functions separate from the applications and services that use these functions. However, network operators and especially internet service providers appear to be becoming more content aware. Networking seems to be becoming infused with content thinking. This might have implications for tomorrow’s internet. To make this point, I will outline three cases that expose the rise of content thinking in seemingly unrelated spheres. But before doing so, I will begin with some contextual information on the architecture of the internet and how it conditions innovation.

The internet architecture keeps networking and content separate

By design, the internet is a general purpose network. This means that the network itself is not tailored to specific uses and applications. Instead, the internet offers a basic networking service upon which complex applications and services can be built. This idea is reflected in the internet’s layered architecture: the lower layers – the internet’s core with the Internet Protocol – include only the essential functions that all applications need. More specific functions can be implemented on higher layers in the protocol stack. For example, the core of the internet does not guarantee low latency or full delivery of all data packets. Whereas low latency may be crucial for multi-player online gaming, other applications such as email or file transfer do not depend on it – so it is not one of the core features. Similarly, preventing packet loss is crucial in online banking, but a little bit of jitter does not make video streaming fail. So the Internet Protocol does not offer any delivery guarantees or optimisation for specific uses or content. For some, this is a bug because the internet does not support their applications in the best possible way. For others, it is a feature, because the internet’s generality makes it more open to innovation at the upper application and service layers.

The internet is open to innovation, but not equally at every layer

The internet’s character as a general purpose network has also conditioned infrastructure development. Most visibly, separating between carriage and content has allowed for these two spheres of internet provision to develop independently. By design, internet carriers do not control what their networking and transport services are used for. The technical architecture limits their role to networking and transport. This is in stark contrast to traditional communication networks such as the telephone or cable systems, where networking and the application service (“voice” or “TV”) were integrated and were offered by the same provider. On the internet, the providers of content services need to develop their applications in such a way as to allow it to be processed in the form and with the features of the Internet Protocol. But they should not need to negotiate with internet service providers about the details of their offerings to end-users. They are free to innovate. Network operators are also free to innovate, but it is much harder to introduce change at the network level. They have to deal with the installed base of the internet. If a carrier came up with a superior networking standard, the chances that it would propagate to the whole internet are small. Look to IPv6 as an example. This successor to the Internet Protocol (v4) was introduced in 1998 (RFC 2460), but even today only an estimated 20 percent of the top 1000 websites can be reached through this protocol. Innovation is difficult to do at the core of the network. A senior network engineer once told me why she thought that nowadays only few young networkers enter the industry: “It is because the cool stuff happens on the application layer!” So, to detect sites of innovation, we can look to where the talent is going.

By design, the perspectives on innovation vary dramatically between internet service providers on the one hand and content & application service providers on the other hand. Content and application service providers may be restricted vertically by the lack of guarantees that the Internet Protocol offers, but the carrier’s ability to innovate is restricted horizontally by the borders between networks. If a carrier wants to introduce, for example, quality of service guarantees, it can only do so at arm’s length by negotiating contracts with interconnected networks. Some carriers have entered such arrangements with each other, but individual arrangements do not scale well. Innovation is a severe challenge when you are in the transport and connectivity business. So it does not come as a surprise that internet service providers struggle with their role as mere enablers of what is perceived as the content provider’s fortune.

The turn to content

What we observe – maybe also as a reaction to the tensions described above – is a turn to content. This may not sound like news, considering that content delivery networks (CDN) have been changing the nature of traffic delivery for more than a decade now. According to Labovitz (PDF) an estimated 50 percent of all internet traffic in the US was exchanged via CDNs in 2013, leading him to conclude: “CDN is the network today.” But generally speaking, CDNs are implemented as so-called overlay networks, which means that their functionality is provided through the upper layers of the protocol stack. However, what I would like to emphasize here is that a content focus is now also emerging around the lower networking layer. Internet service providers have become more content-aware. Three examples from different spheres indicate this.

Case 1: Vertical integration

Just a few days ago, two stories made the news in the networking world. Reportedly, cloud computing and online video service provider Amazon is planning to sell internet access in Europe. Vice versa, internet access provider AT&T announced that it would buy media giant Time Warner. In both cases, carriage and content are being married through vertical integration. This is like reversing the architectural separation between the two by means of economic organisation. Vertical integration puts network operators back in the driver seat in the sense that they gain control over specific content in their network, e.g., it allows for the bundling of services and offers. It also incentivises such meta-companies to treat their own content preferentially, which raises concerns over questions of net neutrality. Further – disregarding where you stand politically – what if Time Warner decided to offer it’s popular content through the AT&T network only? With media companies of this scale, it would change the public sphere.

Case 2: Intelligence in the network

Another news story: The US telecommunications regulator FCC recently passed new rules limiting how internet service providers could use and sell customer data. This ruling has rightly been interpreted in the context of privacy. But it also points to the issue that network operators are on the lookout for new sources of revenue. Some are turning to the content and application services that they are mandated to transmit by augmenting their network with intelligence. This way, they see what their services are being used for, sometimes even by whom. In such scenarios and where regulations allow it, ISPs can become data brokers, or they can even modify the content by injecting ads.

Case 3: Information-centric networking

Finally, the internet infrastructure’s turn to content is being accompanied by the rise of an alternative networking paradigm that emphasises content. It is called information-centric networking and has attracted internet researchers since the early 2000s. The idea behind information-centric networking is to make named data addressable and routable. So instead of providing host-to-host connectivity that relies on IP addresses and the end-to-end principle, in an information-centric networking architecture information objects would be requested and delivered. Internet routing would shift from “where” to “what”, from location to identity, underlining the network operator’s awareness of content. It is unclear whether information-centric networking will mature and move beyond its current application sites in test beds and isolated networks. But some people are certainly hoping for it to become the “new waist of the hourglass as the basis of a reincarnation of the Internet (sic!)” (RFC 7476).

A good 20 years after the commercial internet began in 1995, the tension between internet service and content is more relevant than ever. APNIC’s chief scientist Geoff Huston even contemplated about the possible “death of transit”, a specific type of internet service. Complementing this, I have tried to delineate further phenomena that indicate this turn to content. Now, of course one could argue that carriers should not lament the uncertainty of their business outlook. Nobody kept them from innovating on the upper layers of the protocol stack. That is true. But it would be missing the point. Because another truth is: the internet as a whole is only as accessible, open and innovative as the underlying network allows it to be.

The internet infrastructure’s turn to content puts some “internet givens” up for discussion. In the near term, there is the question of global connectivity and the role of carriers as providers of this basic internet service. If it holds true that their business model is eroding, are we willing to accept changes such as more companies that jointly own media and networks or more “intelligence” in the networks? Or is it time to think about societal backup systems which ensure that anyone can communicate with anybody else on the internet? Because in the original internet architecture the principles of reachability and end-to-end connectivity are paramount, but they are not so in content-driven networking – in whatever way it is realised. In the medium run, we should look openly at information-centric architectures as alternative paradigms. This article has taken a critical stance so far, but it needs to be acknowledged that such architectures clearly address the heterogeneous needs of those who use the internet infrastructure. That said, let’s be careful what we wish for. As has been recently documented by the Internet Research Task Force group on information-centric architecture, new policy challenges are arising e.g., in the fields of access control, encryption and at the points of passage between virtual identities and “real-world” identities. So the infrastructure’s turn to content in itself is neither good nor bad; but it should make us care about the future of the network. And that is a choice, not a given.

Thanks to Kirsten Gollatz, Sebastian Möhrenfeld and Ashwin Mathew for feedback.