THE INTERNET AS A LAYERED, END-TO-END PLATFORM
The Internet is a uniquely vast emergent system - a platform for untold forms of economic, social, and personal interaction. Holland and Miller note that ‘complex adaptive systems usually operate far from a global optimum or attractor.
Such systems exhibit many levels of aggregation, organization, and interaction, each level having its own time scale and characteristic behavior’ (1991, p. 365). The Internet’s technical and organizational architecture exhibits these types of distinct levels, and its layered structure is imbued with a set of guiding principles (Solum and Chung, 2003). This section discusses four fundamental architectural principles inherent in the Internet - principles that make it not only more complex, but also more likely to facilitate productive adaptation and experimentation. Engineers have created the Internet’s architecture collaboratively over many years, with an eye to these principles. This architecture is what gives the Internet its exceptional value and ability to promote growth (Whitt, 2013).The technical architecture of the Internet is at once simple and complex. The relatively simple Internet Protocol (IP) sits in the ‘middle’ of all Internet communications. Any device or application that connects to the Internet will ultimately have its communications translated into this lingua franca of the network. However, IP is highly extensible because it allows for new virtual protocols to be built ‘above’ its layer, or for new physical devices to implement ways to transport IP traffic ‘below’ its layer. Higher-layer protocols include HTTP (a standard for web traffic), and lower-layer protocols include Ethernet (a standard for wired devices). This is what makes up the so-called ‘hourglass’ structure of the Internet. As Bush and Meyer (2002) explain:
In this model, the thin waist of the hourglass is envisioned as the (minimalist) IP layer, and any additional complexity is added above the IP layer.
In short, the complexity of the Internet belongs at the edges, and the IP layer of the Internet should remain as simple as possible.Organizationally, there is no single governing body or process that directs the development of the Internet’s architecture. Instead, we have multiple bodies and processes of consensus. Much of the ‘governance’ of the Internet is carried out by so-called multistakeholder organizations, such as the Internet Society (ISOC), the World Wide Web Consortium (W3C), and the Internet Corporation for Assigned Names and Numbers (ICANN) (Sieh and Hatfield, 2012). Over the last two decades, although these entities have largely established the norms and standards for the global Internet, ‘they are little known to the general public, and even to most regulators and legislators’ (Waz and Weiser, 2012, p. 1). The IETF operates under the auspices of ISOC, and its stated goal is ‘to make the Internet work better’ (Alvestrand, 2004, p. 1). It is the institution that has developed the core networking protocols for the Internet, including IP, TCP, UDP (User Datagram Protocol), and countless others. The IETF is open to any interested individual, and conducts activities through working groups in various technical areas.
The IETF Requests for Comments (RFCs) were first established in April 1969 by Steve Crocker at UCLA. The memos were intended as an informal means of distributing shared ideas among network researchers on the ARPANET project, the precursor to the Internet. ‘The effect of the RFCs was to create a positive feedback loop, so ideas or proposals presented in one RFC would trigger other RFCs’ (Leiner et al., 1997). Once consensus came together within the IETF, a specification document would be created in order to serve as the basis for implementation by various research teams. Many RFCs have become de facto standards for core Internet Protocols. It is well established that:
Internet Protocols not only serve critical technical functions but can have significant political and economic implications...
The open publication of Internet standards with minimal intellectual property restrictions has enabled rapid innovation and has generally produced the market effect of full competition among companies developing products based on those standards. (DeNardis, 2010, p. 7)Many scholars have attempted to articulate the principles embedded in the Internet’s architecture (Bernbom, 2000; Barwolff, 2010; van Schewick, 2010; Doria, 2011). RFC 1958 provides perhaps the best workable foundation: ‘In very general terms, the community believes that the goal is connectivity, the tool is the Internet Protocol, and the intelligence is end to end rather than hidden in the network’ (Carpenter, 1996). It is likewise clear that ‘modularity’ - or layering - is the logical scaffolding that makes it all work together.
3.4.1 Modularity
The use of layering means that functional tasks are divided up and assigned to different software-based protocol levels. For example, the ‘physical’ layers of the network govern how electrical signals are carried over physical wiring. Independently, the ‘transport’ layers deal with how data packets are routed to their correct destinations, and what they look like. The ‘application’ layers control how those packets are used by an email program, web browser, or other user application or service.
This simple and flexible system creates a network of modular ‘building blocks’ that can be adapted and evolved. Applications or protocols at higher layers can be developed or modified with little impact on lower layers, while lower layers can adopt new transmission and switching technologies without requiring changes to upper layers. Innovations at each layer are thus unconstrained and independent, while at the same time being interdependent on the functions of the other layers. Therefore, it is important that there exist stable interfaces between the layers, and that a particular bug or feature be addressed at the appropriate layer. Computer engineers call this type of segmented functionality ‘abstraction’.
In the case of the Internet, this approach has facilitated tremendous experimentation and innovation throughout the system.The Internet succeeded where other attempts at large-scale networks failed. Not only did the Internet overtake commercial ventures like AOL, CompuServe, and Prodigy, it proved far more amenable to adaptation than similarly situated governmental initia- fives. By the early 1980s, the French government had launched a project called ‘Minitel’, which sought to establish a nationally networked system of personal computers. Each device would call in to a set of centralized ‘Kiosques’, while France Telecom billed for usage. Five million Minitel units were distributed free of charge to citizens and businesses, and the system became pervasive in French society. However, as Alcouffe and Alcouffe (2009, p. 213) note:
As successful as the widespread acceptance of the Minitel was, it effectively set a new technological standard that was not open to competitors. This locked in a technology that failed to keep up with other technological developments such as packet-switching protocols (e.g. TCP/IP) and protocols that are broadly interoperable across a wide variety of communication service providers with other network providers.
In the language of complexity economics, Minitel simply did not allow enough evolutionary experimentation. In the language of Internet engineering, it was unitary and monolithic rather than modular and layered. By the 1990s Minitel was rapidly losing ground to the Internet, and it was finally shut down on 30 June 2012.
In all engineering-based models of the Internet, the fundamental point is that the horizontal layers, defined by code or software, serve as the functional components of an end-to-end communications system (Whitt, 2004).
3.4.2 End-to-end
RFC 1958 states that ‘the intelligence is end to end rather than hidden in the network’ with most work ‘done at the fringes’. The function of the middle portion of the IP hourglass is simply to deliver communications - without processing, changing, or discriminating between traffic.
This ‘end-to-end principle’ arose in the academic communities of the 1960s and 1970s. It took hold when the US government compelled adoption of the TCP/IP, mandated a regulated separation of conduit and content, and granted non- discriminatory network access to computer device manufacturers and dial-up online companies (Whitt, 2009, pp. 507-8). Consequently, these end-to-end arguments ‘have over time come to be widely considered the defining if vague normative principle to govern the Internet’ (Barwolff, 2010, p. 134).End-to-end tells us where to place the network functions within a layered architecture. Layers that are closer to the ‘middle’ of the network should serve merely to connect, rather than to make decisions on behalf of the nodes. The Internet has thrived as a complex system because the agents using the system can rely on the network to connect them without hindrance. This maximizes their evolutionary options, and increases the degree of overall emergence.
The end-to-end principle is reasonably straightforward to apply at the core of the network, even when some reasonable exceptions can be made for firewalls and traffic shaping (Whitt, 2009, p. 453, n. 199). Moving up to the application and content layers, there is considerably more discretion. Application developers and content producers are themselves adaptive agents. Thus, the end-to-end principle has become a general mandate to make the basic Internet Protocols simple, general, and open, leaving room for innovation at higher layers.
3.4.3 Interconnection
RFC 1958 puts it plainly: the goal of the Internet is ‘connectivity’. Unlike the earlier ARPANET, the Internet is a collection of IP networks owned and operated by private telecommunications companies, governments, universities, individuals, and others - each of which needs to connect with others. Kevin Werbach has pointed out that, ‘the defining characteristic of the Net is not the absence of discrimination, but a relentless commitment to interconnectivity’ (Werbach, 2007, p.
1273). Jim Speta agrees that the Internet’s utility largely depends on, ‘the principle of universal interconnectivity, both as a technical and as an economic matter’ (Speta, 2003, p. 17). The early Internet was designed with an emphasis on internetworking and interconnectivity, and moving packets of data transparently across a network of networks. Steve Crocker reports that even in the preInternet environment all hosts would benefit from interconnecting, but that ‘the interconnection had to treat all of the networks with equal status’ with ‘none subservient to any other’ (Crocker, 2012).Today’s Internet embodies a key underlying technical idea: open-architecture networking. Under this design principle, network providers can freely interwork with other networks through ‘a meta-level “internetworking architecture”’ (Leiner et al., 1997, p. 103). Critical ground rules include that each distinct network must stand on its own, communications will be on a best-effort basis, and there is no global control at the operations level.
The interconnecting of disparate networks has direct economic implications. ‘Interconnection agreements do not just route traffic in the Internet, they route money’ (Clark et al., 2011, p. 2). Likewise, a healthy flow of money from end users to Internet service providers (ISPs) - absent market power abuses by the ISPs - is essential for sustained infrastructure investment.
More broadly, it is clear that the Internet’s interconnection principle supports the observation that complex systems have the greatest opportunity to grow when they are highly networked. Agents have more avenues for collaboration, learning, and adaptation as their universe grows beyond their local network. A densely connected set of nodes has more chances for novel innovation, and greater means to ‘route around’ market failures.
3.4.4 Agnosticism
RFC 1958 states that, in order to achieve connectivity, ‘the tool is the Internet Protocol’. By design, IP is completely indifferent to both the underlying physical networks, and to the countless applications and devices using those networks. In particular, IP does not care what underlying transport is used (such as fiber, copper, cable, or radio waves), what application it is carrying (such as browsers, email, instant messaging, or MP3 packets), or what content it is carrying (text, speech, music, pictures, or video). Thus, IP enables any and all user applications and content. This enables incredible diversity of use. ‘The system has standards at one layer (homogeneity) and diversity in the ways that ordinary people care about (heterogeneity)’ (Palfrey and Gasser, 2012, p. 108).
RFC 172, discussing the File Transfer Protocol, says that the network should ‘assume nothing about the information and treat it as a bit stream... whose interpretation is left to a higher level process, or a user’ (Bhushan et al., 1971, p. 5). As Barwolff puts it, IP creates ‘the spanning layer’ that creates ‘an irreducibly minimal coupling between the functions above and below itself’ (2010, p. 136). Not only does IP separate the communications peers at either end of the network, it generally maintains a firm separation between the entities above and below it.
As a result of this core agnosticism, the Internet - like other highly complex systems - can generate surprising outcomes. As one neo-Schumpeterian puts it, ‘The contemporary technological paradigm generates novelty, uncertainty, and surprise in an unprecedented way’ (Dopfer, 1994, p. 159). Economists and policy-makers should embrace this agnosticism, rather than trying to predict or impose specific outcomes on the system.
We must keep in mind that these four principles describe the Internet in its native environment, with no alterations or impediments imposed by other agents in the larger ecosystem. Where laws, regulations, or other activities would curtail these design attributes, the Internet may become less than the sum of its parts. It is only when the design features are able to work together that we see the full emergent phenomenon of the Internet (Whitt, 2013).
3.5