Data, Geopolitics, and the Governance of Cyberspace
Kieron O’Hara and Wendy Hall
1
Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and certain other countries.
Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by license, or under terms agreed with the appropriate reproduction rights organization. Inquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above.
You must not circulate this work in any other form and you must impose this same condition on any acquirer.
Library of Congress Cataloging-in-Publication Data
Title: Four internets : data, geopolitics, and the governance of cyberspace / Kieron O’Hara and Wendy Hall.
Description: New York, NY, United States of America : Oxford University Press, [2021] | Includes bibliographical references and index.
Identifiers: LCCN 2021014363 | ISBN 9780197523681 (hb) | ISBN 9780197523704 (epub) | ISBN 9780197523711 (Online)
Subjects: LCSH: Internet governance. | Internet--Political aspects. Classification: LCC TK5105.8854 .O43 2021 | DDC 384.3/34--dc23 LC record available at https://lccn.loc.gov/2021014363
DOI: 10.1093/oso/9780197523681.001.0001
9 8 7 6 5 4 3 2 1
Printed by Sheridan Books, Inc., United States of America
CONTENTS
Foreword vii
Preface xi
Acknowledgements xv
1. Preliminary Concepts: Networks and Data 1
PART I: The First Internet: The Silicon Valley Open Internet
2. How the Internet Developed 27
3. Governing the Internet 36
4. The Vision of the Open Internet 51
5. Policy Question: How Can Quality Be Ensured in an Open System Like Wikipedia? 59
PART II: Alternatives to Openness
6. Openness and Its Discontents 71
7. The Second Internet: The Brussels Bourgeois Internet 77
8. Policy Question: When Is Surveillance Justified? 92
9. The Third Internet: The DC Commercial Internet 103
10. Policy Question: How Can Competition against the Tech Giants Be Fostered? 117
11. The Fourth Internet: The Beijing Paternal Internet 125
12. Policy Question: Is Huawei Infrastructure a Threat to Western National Security? 145
13. The Moscow Spoiler Model 154
14. Policy Question: Is a Sovereign Internet Feasible? 173
15. Peaceful Coexistence 177
PART III: Futures
16. India, the ‘Swing State’ 183
17. Policy Question: When Should Personal Data Cross Borders? 200
18. Artificial Intelligence 206
19. Smart Cities and the Internet of Things 218
20. Social Machines 229
21. The Unity of Freedom 237
Glossary of Abbreviations 247
Notes 251
References to Books and Academic Papers 293 Index 321
FOREWORD
Predicting the future is never an easy task, given its inherent uncertainty. Old friends Wendy Hall and Kieron O’Hara have made as good a shot as possible by employing the tactic of exploring several possible futures based on experience and existing trends. We have examples of the Internet’s manifestation in different geo-political settings and these have informed the four Internets envisioned in this thoughtful book. Among my favourite personal life theorems is #206: ‘Everything is more complicated’. I think that may be the case here, as well.
The original Internet design envisioned a diverse collection of packetswitched networks held together by ‘gateways’ and a common set of protocols, starting with the Internet Protocol (IP) layer, which included a global addressing scheme that would locate a ‘host’ computer as to which network it was in and where it was attached to that network. The hosts and the gateways would be IP-aware although the underlying networks themselves did not need to be. In fact, in the original experimental implementation, the Arpanet, Packet Radio net, and Packet Satellite net were entirely unaware of the overlay Internet. Gateways between the networks encapsulated and decapsulated Internet Packets into and from the payloads of packets in the underlying networks. Internet packets were like postcards placed inside of envelopes for transport within a constituent network. Networks were not limited in geographic scope and could be local or global. The addressing structure was topological, unlike the country- coded telephone network.
It was assumed that any host could address an Internet packet to any other host in the system, implying universal addressability. This was considered essential because it was not clear from the outset which hosts might need to communicate with which other hosts, nor was it clear what applications might be needed or invented. A consequence of the latter idea is that the Internet packets had no information as to how they were transported— whether they were going over twisted pair copper, mobile radio, satellite, Ethernet, or some other bearer. Thus, when optical fibre appeared in the early 1980s, the
Internet packets were happily transported that way, without any change to the basic Internet architecture.
Moreover, the Internet packets were unaware of the applications they supported. From the protocol point of view, they simply were carrying a payload of uninterpreted bits to be delivered to the destination host computer. The host computer would interpret the payload to implement the application. A packet might contain part of an email message, bits of video or audio, a fragment of an image, part of a spreadsheet, and, later, portions of Web pages. This lack of application awareness was crucial to the Internet’s rapid evolution of applications because the architecture of the basic Internet did not have to change to support new applications. The so- called end- to-end transparency of the Internet has served the application developers well. Implementers needed neither permission nor alteration of the Internet to invent a new application, with the possible exception that some applications required more transport capacity to be feasible. Thus it was that streaming audio, video, and videoconferencing required a substantial increase in transport bandwidth, mostly provided by optical fibre and more recent developments in high speed radio transmission, before they could be widely implemented.
The Internet architecture is layered with well-defined interfaces between the layers. These interfaces have been largely stable since the beginning of the Internet’s design and this has been beneficial for two reasons. The first is that the lower interface of a new protocol benefits from being able to rely on the stability of its interface to the layer below on which functionality it depends. On the other hand, the new protocol may exhibit a new upper interface and functionality in support of new applications. So, for example, a RealTime Protocol (RTP) may rely on the lower, unreliable User Datagram Protocol (UDP) in support of real- time speech. UDP does not guarantee orderly and complete delivery, unlike the Transmission Control Protocol (TCP), but it tries to be low latency. RTP might not re-transmit lost packets and might discard ‘old’ packets but otherwise might try to deliver packets in time sequence order for a speech application.
Implicit in the end- to-end and universal connectivity assumption is essentially free flow of data across geo-political borders. That assumption is vital, for example, for the implementation of multi-national ‘cloud’ computing systems in which data centres exchange data freely for purposes of replication to prevent loss of data. Various forms of fragmentation occur when this free flow is impeded, blocked, or filtered in some fashion. It is also important to recognize that the freedom of interchange can be eroded at different layers in the architecture. For example, a country might require that certain domain names be blocked by interfering with domain name lookups in the Domain Name System. Internet users in China cannot reach the domain www.google. com because the lookups return ‘no such host’. The Chinese government can
[ viii ] Foreword
implement this effect by controlling the behaviour of the in- country domain name ‘resolvers’. Attempts to bypass this control using ‘virtual private networks’ may work but can be detected and blocked.
Fragmentation of the Internet can occur at other layers in the system. Indeed, some countries require that search engines such as Google and Bing must not show responses for certain queries. In Germany and France, for example, it is illegal to facilitate any pecuniary benefit from the sale of Nazi war memorabilia from World War II. The European ‘Right to be Forgotten’ requires that search engines maintain a list of name-based lookups, the response to which must be discarded or blocked. Most countries require that Internet providers or application providers block the transport of child pornography or other child-based exploitation content. There can be harsh penalties for failure to meet these standards. Intellectual property violations (usually, copyright) can also engender serious fines and penalties which vary from country to country.
The general point is that discovery, retrieval, and transport of certain kinds or sources of information or data are prohibited in various ways across the global Internet. All of this adds up to varying degrees of interference with the free flow of information in the Internet. In many cases, such constraints are broadly agreed among most countries that house portions of the Internet. But there is also disagreement with some restrictions on the grounds that they violate universal rights of speech, access to information, ability to assemble (if virtually), and so on.
In this book, the authors select four possible representative futures for the Internet’s evolution, depending on geo-political and policy frames that affect the otherwise free flow of information in the network. In reality, there already are and will be nuanced demands made by governments of the Internet access and application service providers and some of these demands will have extraterritorial character. The European Right to be Forgotten and the General Data Protection Regulation have already manifested outside of the European Union. The notion of ‘Data Sovereignty’ is becoming a popular meme, attempting to mirror in cyberspace the concepts captured in the 1648 Peace of Westphalia that countenanced sovereignty within geo-political boundaries. I am unpersuaded that this notion is beneficial, particularly in the context of distributed cloud computing which benefits everyone through the replication of data at multiple data centres to reduce latency of access (e.g. Content Distribution Networks) or to protect against data loss.
The arrival of social networking in the form of Facebook, Instagram, YouTube, Twitter, TikTok, WeChat, among many other applications, has ushered in new hazards to the free flow of information in the Internet. Of course, the issues here are misinformation, disinformation, hate speech, bullying, fake news, and a bevy of other speech-related concerns, to say nothing of the
x apparent feedback reward mechanisms that appear to exacerbate extreme expression for the purpose of gaining attention and potentially income from shared advertising revenues. To this we can also add the general problems of malware including denial of service attacks, hacking, ransomware, identity theft, webpage defacement, and other harmful or disruptive behaviours. That these can pose serious threats to infrastructure such as the power generation and distribution systems, healthcare systems, communication systems, and transportation control systems underlines the gravity of responses necessary. Whether such attacks rise to the level of national security and even military response is a conundrum for our twenty- first century. All of these potential hazards drive countries to adopt policies in response which may not be compatible in bilateral or multilateral terms. Indeed, these issues have elicited lengthy and unresolved debates over how they should be resolved and whether they can be treated through multilateral conventions and treaties akin to the Law of the Sea or agreements as to the peaceful uses of space.
In the long term, I hold the view that the Internet’s benefits are wellestablished but that there are risks inherent in its abuse. Our task is to understand and articulate the various potential paths forward and to do our best to assess their implications, as Wendy Hall and Kieron O’Hara have attempted here.
8 November 2020
Vinton G. Cerf Vice President and Chief Internet Evangelist Google, LLC
PREFACE
The Internet appears a fixed part of modern reality. Its roots date back over fifty years, and by the first quarter of the twenty- first century it has become the pre-eminent means of distributing information. To those in their early twenties, it will have been a permanent, if mutable, background to their lives. It seems unquestionable, like electricity and roads. Its design was intended to make it tolerant of disruption and faults, and it is pretty robust and resilient. We can go further: it is surely evident by now (this book was completed in lockdown during the 2020 COVID-19 pandemic) that it has become a piece of critical infrastructure, and needs to be protected as such.
Yet it is surprisingly delicate all the same, a gossamer arrangement of different types of hardware, protocols to describe how systems communicate with each other, and methods of social coordination, ranging from regulations to contracts to what amount to little more than gentlemen’s (and ladies’) agreements. The Internet is not a monolithic technological creation, patented and run by a single company or government, but a congeries of systems, protocols, standards, hardware, and organizations. Some of these organizations have national standing, some have global reach, and others have international standing. Some are public bodies, some private companies, and some non-profit organizations. The Internet is one of the few institutions where the people round the table are as likely to be in corporate suits and Christian Louboutins, as black t- shirts with hoodies and Air Monarchs, or even (still, just about) tie-dyed denim and bare feet.1
Furthermore, the system is truly sociotechnical— we cannot hive the technology off from the people who use it in their everyday lives. Every design decision reflects, and imposes (perhaps unconsciously), a balance of power, while cultural, economic, and political tensions play out across collective-action problems. Neither computer science nor the social sciences are individually sufficient to encompass all the study required to understand the most complex piece of technology ever created, the structure of which is driven by the people who upload, download, and link content. The authors of this book have
long argued that concentrated interdisciplinary research, encompassing social and technical studies, is required both to understand it and to engineer it.2
The Internet evolves so quickly that there is no good time to write about it. The year 2020 was peculiarly inopportune, it appears to us (now we have done it). The uncertainties were immense, and prediction seemed hopeless. We were privileged to be asked to write a short piece for a collection of prognostications for the future published in December 2019 called The Wired World in 2020. Not a single author predicted that we would spend most of the year sitting indoors while over a million people died, and the world economy tanked.
The future of the Internet depends on answers to many questions (some of which will by now be evident to the reader). Most obviously, the COVID-19 pandemic has completely reshaped our relationships with the Internet, and with the wider world. What will the new normal be?
There are political questions, of which the elephant in the room during the time of writing was the outcome of the 2020 US presidential election. President Trump’s success or failure in 2020 would have massive influence over the Internet’s future, and that of the technology industry generally. President Biden will certainly govern in a different style, but there may well be aspects of his predecessor’s agenda that he preserves. Other unresolved issues concern the extent to which China will challenge the United States for global leadership (it currently heads four of the UN’s fifteen specialized agencies, including the International Telecommunication Union), and how the European Union will work through Brexit, and the post-Merkel era.
The Internet is a co- creation of its users, and every year it adds in excess of a hundred million new co- creators. These people will change the Internet in unpredictable ways. Our study of India shows some of the different directions these changes may take.
Finally, there is the technology. The Internet creates data, the fuel for artificial intelligence and smart cities, which many see as the future for humankind. Will they produce utopia, or dystopia, or are they merely hype? Other important technological trends we had to ignore or skate over for reasons of space are blockchain, automatic face recognition, virtual or augmented reality, undetectable deepfake technology, quantum computing, ubiquitous wireless broadband, and medical wearables. Any one of these could have a massive effect on our privacy, public life, and well-being.
Our research, as we describe in Chapter 1, is structured by three basic ideas about the Internet as a network, the Internet as a producer of data, and the Internet as the key technology for digital modernity.
• The different networks connected by Internet protocols are governed by subtly different principles, which may, over time, make it harder to get data
across the Internet as a whole, and therefore may increase the pressures to fragment.
• These different principles have a strong connection with attitudes towards data, which is the chief source of power on the Internet.
• The visions and models of the Internet we discuss in this book are all, in their way, responses to the prime imperatives of digital modernity.
We hope to weave the narratives of networks, data, and modernity into an understanding of how different cultures and ideologies see value in the Internet, and how it might be possible to keep it together, while respecting difference and diversity. The key, we claim, is how the Internet is governed.
The book is divided into three parts. In Part I, we introduce the question of Internet governance, and how it reflects values. We set out the original Internet vision, the ultimate techie ambition of an open, unmanaged, free information space.
In Part II, we consider some of the problems with this vision, and then look at the alternatives that have evolved through time. We focus on three that have become highly influential; these, plus the original open vision, give us the four Internets of the book’s title. These aren’t the only visions that are available— hence our title is Four Internets, not The Four Internets; a definite article would be misleading. Other Internets could be imagined, if the appropriate ethical vision found a technological realization and sufficiently powerful institutional backing. We will also look at another ideological model of how the Internet can be used as an instrument of policy—not one that comes with its own positive vision, but a spoiler that might undermine the integrity of the Internet and thereby (perhaps accidentally) contribute to fragmentation.
Part III of the book sets out some of the implications of the ideological centrifugal force. This isn’t a Jeremiad, but it seems clear that the more closely the diversity of Internet governance tracks the diversity of the world it spans, the easier it will be to contain this force. There are radically different views of how communications should be managed, and these need to be reflected and respected in governance structures—even if engineering imperatives are compromised as a result.
Each geopolitical entity we describe is at the centre of many policy wrangles. To illustrate some of the choices and trade-offs involved in these, each of the six geopolitical chapters is succeeded by an unresolved ‘policy question’, giving us a deeper dive into how politics, technology, and regulation converge and diverge.
ACKNOWLEDGEMENTS
Conversations with a great many colleagues have contributed to the work you have before you, and our debts are immense. For direct help with this research, many thanks are due to Paul Smart for many ideas, and in particular his research into the Chinese AI and data protection scene. Thanks also to Les Carr for managing relevant projects and many illuminating discussions. Special thanks to Anni Rowland- Campbell for brave conversations, comments, and references. Our colleagues past and present at the University of Southampton Web Science Institute, drawn from across the university’s various faculties, have been a constant source of inspiration, reminding us of the sheer range of perspectives from which one can view the Internet. The Web Science students we have taught over the years have—as is always the case— taught us at least as much in return.
We are particularly grateful to those brave and patient souls who read the whole thing and commented copiously: Peter Kingsley, Carolyn Nguyen, Srinath Srinivasa, Phil Tetlow, Bill Thompson. Thanks in particular to Vint Cerf, who not only commented, but kindly agreed to write our Foreword. Thanks also to anonymous reviewers at Oxford University Press, both at proposal and submission stage, and to Sarah Humphreville and the team at OUP who piloted us through the process. The feedback we received was extremely helpful, and it goes without saying— though we will now say it— that any errors that remain are our responsibility solely.
We are grateful to Mark Schueler and Jie Tang for their diagrams depicting the growth of the Web.
The ideas first saw the light of day as we collaborated to write a paper for the Centre for International Governance Innovation. Our gratitude goes to them for the opportunity, and also all the help of Fen Hampson and their excellent team in refining the argument, and presenting the paper so that we started to get feedback from a range of communities and disciplines. The full set of papers by us in which the broad argument has been presented (all authored by O’Hara & Hall, except where stated) is:
• Four Internets: The Geopolitics of Digital Governance, CIGI paper no. 206, https:// www.cigionline.org/ publications/ four- internets- geopoliticsdigital-governance
• ‘There are now four competing visions of the internet. How should they be governed?’ World Economic Forum Agenda blog, https://www.weforum.org/ agenda/ 2019/ 03/ there- are- now- four- competing- visions- of- theinternet/, re-published in Global Policy, https://www.globalpolicyjournal. com/ blog/ 14/ 03/ 2019/ there- are- now- four- competing- visions- internethow- should- they-be-governed
• ‘The dream of a global internet is edging towards destruction’, The Wired World in 2020, https://www.wired.co.uk/article/internet- fragmentation
• ‘Four Internets’, Communications of the ACM, 63(3), 28–30, https://doi.org/ 10.1145/3341722
Thanks to audiences to which the argument has been presented, in particular the techies at the Our People- Centered Digital Future Conference in San Jose, California, an enthusiastic audience of young Chevening Scholars at the University of Southampton, a somewhat more mature audience at the University of Luxembourg’s 2019 Rentrée Académique, and a mix of Parliamentarians and industrial technologists at the 2020 Parliament and Internet conference in Westminster.
Thanks also to journalists who have aired the arguments and given us a sense of how they may be received, including Martin Sandbu of the Financial Times, Jordi Peréz Colomé of El País, and Amol Rajan of the BBC.
[ xvi ] Acknowledgements
CHAPTER
1
Preliminary Concepts
Networks and Data
The Internet is a technical system, but not purely a technical system. That it is a technical system brings with it a vital invariant that can never be neglected: it has to work. If it does not actually perform its function, then the question of governance may be of academic interest but is hardly pressing. But because the Internet is more than just a technical system, then it will inevitably be measured against other values concerning its role in society. The point of governance is to ensure that the system functions, while respecting other values too.
Evaluating that respect depends on many aspects that have become more complex as the Internet has grown. It is a global resource, used for commerce, politics, entertainment, health and well-being, socializing, scientific research, and education. It sustains the most advanced economies, and at the same time is a tool for development. It is a voice for the excluded, and a megaphone for the powerful. It is also a cog in many complex systems, delivering government services, hosting cryptocurrencies, coordinating defence and weapons systems, and enabling emergency services. It generates a record of our activities that can be used to personalize the services we receive, or that can be held against us in the future. It is the sine qua non for so many valuable things that hard choices, dilemmas, and trade-offs seem inevitable.
It is a global Internet. This means, in our ineffably plural world, that many different points of view must be taken with respect to its governance. What seems self-evident to a professor of computer science and fan of the Grateful Dead in San Francisco might be somewhat less obvious to a businessperson who relies on it to coordinate a complex supply chain across East Asia that
employs thousands of people, and positively counterintuitive to a law enforcement officer in a combustible part of the world whose main job is containing the threat of inter- community violence. Different people have to cooperate, and agree to differ. As economist Jonathan Cave puts it:
There is broad support among experts and laymen alike for a regulatory system based on common and attractive architectural and policy principles and values, many of which are endorsed throughout . . . currently-dominant developed nations. But they are not universally supported in the developing world and do not in any case receive the same ranking.
If ‘our’ future internet differs widely from the global internet or the internets of other global powers, we may lose purchase or find ourselves bypassed. Expressing our values and reaching our objectives may be frustrated or excessively costly.1
In this book, our aim is to explore some of these issues, and to consider some of the value- sets that have emerged, with a boost from powerful geopolitical actors. We will not— cannot—decide how the Internet could, should, or must be. We, the authors, are as embedded in particular value systems as anyone else, and consequently as biased. We will try to take a dispassionate view of the range of perspectives, and think about the pressures they create. We will survey the powerful emerging visions of the Internet, show how they are implemented, and discover what happens when they clash. We will try to understand how new and exciting innovations in Internet technology may be shaped, depending on which of these visions gains a no doubt temporary upper hand. And ultimately, as these ideological conflicts are here to stay, we want to understand how Internet governance can best accommodate these heterogeneous views, while keeping the show on the road.
Perhaps most importantly, at some point in 2019, the proportion of people connected to the Internet exceeded 50 percent. This is an amazing feat, for which the Web, social networking, gaming, and streaming services must take a lot of the credit; it is not long since it was customary to point out that half the world’s population had never even made a phone call, and there have long been claims that Internet growth would reach a ceiling. Three and a half billion people are now connected, and the number is only likely to grow. The question then is how the next three and a half billion will change the Internet, for change it they surely will.2
One study based on World Bank data from 2016 found the following percentages; at the time, global penetration was 46 percent, so note how the South Asian and sub- Saharan African regions pulled down the average (Table 1.1).
Growth will not be from the rich world; the United States has a surprisingly low rate of Internet penetration, but no other rich democracy has further
Table 1.1. INTERNET PENETRATION IN DIFFERENT REGIONS AND SELECTED COUNTRIES (2016)
Region Percentage
North America 78
Europe and Central Asia 73
Latin America and Caribbean 57
East Asia and Pacific 53
Middle East and North Africa 48
South Asia 26
Sub- Saharan Africa 20
Highs (%)
Bermuda (98), Canada (90)
Iceland (98), Luxembourg (98)
Lows (%)
United States (76)
Romania, Bulgaria (60; lowest in EU), Bosnia and Herzegovina (55; lowest in Continental Europe), Turkmenistan (18)
Source: Max Roser, Hannah Ritchie, & Esteban Ortiz-Ospina, ‘Internet’, Our World in Data, https://ourworldindata.org/internet, except Bermuda figure from https://www.statista.com/statistics/731257/ bermuda-internet-penetration/
room. The new users are therefore going to be in sub- Saharan Africa, India, rural China, and other parts of South and East Asia. The people in these regions are not so different from everyone else; we must expect that they will be interested in news, sport, entertainment, pornography, dating, and networking,3 but the balance of ideologies will change, and so will the pressures on Internet governance. We can’t predict the exact nature of those pressures, but we can at least prepare for them.
To begin at the beginning, the Internet is a network of computers, connected for instance by copper or fibre-optic cable, which can carry data 4 ‘Internet’ is short for ‘Internetwork’, and implies a network of connected computer networks. What makes it the Internet, as opposed to any old computer network, is the use of two particular sets of communication rules: the Transmission Control Protocol (TCP), and the Internet Protocol (IP), which together are known as the Internet Protocol Suite (TCP/IP). IP defines how computers can exchange data with each other, while TCP governs the connections between computers to exchange data using IP. A system of IP addresses identifies individual computers in such a way as to (a) locate each computer in a specific network, and then (b) identify a route through that network to the computer itself using IP. Between them, these protocols provide an end- to-end specification of how data gets transferred across the Internet. We will describe these protocols (without getting too technical), and their evolution, in Chapter 2.
The Internet is, then, a network of computer networks connected by TCP/ IP, on which data can get from anywhere on the Internet to anywhere else pretty well instantaneously. This is what makes it so valuable—data becomes abundant, and can be reused by many people, sometimes in unanticipated and propitious circumstances. Part of the cleverness of the design lies in the fact that the Internet infrastructure is simultaneously a means of building a network designed to pass data between the nodes, and a means of stitching together pre-existing or independent networks, even when they operate on very different principles from each other.
Our investigations in this book therefore have two parallel foci. The first is the connectivity and size of the network. How are the nodes of the network connected, and how easy is it to get data from one part to another? The second is what is moved through the network facilitated by the connections. The ability to amalgamate data into larger aggregations, and then process it en masse, is the source of great technological potency. This can be done by bringing increasingly large quantities of data together at single points in the network, or by clever techniques using the network to bring inferential power to query data at its various different locations, thereby creating a large virtual dataset.
The value of the Internet lies in its status as a network, and as a data infrastructure. The job of this chapter, therefore, is to sketch the sometimes surprising properties of networks and data, before we proceed to the Internet itself. In particular, in terms of value and power, we will differentiate networks and hierarchies, and in terms of expressive power we will contrast data with information. The effects of the existence of an efficient, global, data- carrying network have upended a great deal of political, economic, and social thought; as our final introductory task, we will also describe the corresponding narrative
of digital modernity as essential background. Where networks provide accessible means by which data can be collected and consumed at scale, many new technological affordances, such as artificial intelligence and smart cities, are created. These affordances, which are at the heart of digital modernity, will be discussed in Part III of this book.
NETWORKS AND THEIR VALUE
A network can be described using a set of points or nodes, with connections or edges between them, to create a graph. An edge connects a pair of nodes, but not every pair of nodes needs to be connected. The connections are usually defined in a particular way— for instance, in a network of human acquaintances, two nodes (people) would be connected if they knew each other, and the edge between them would symbolize their acquaintance. In a network of holding companies, two nodes (companies) would be connected if one of them was the legal owner of shares in the other. In a family tree, which is a type of network, two people would be connected if one was the parent of the other, and the edge would symbolize ancestry. Some connections have a direction (like ownership and parenthood), others (like acquaintance) don’t. If the connections of the network are directed, this is a directed network; if not, it is undirected. If two nodes are not directly connected, there may still be a route between them in the network. For instance, A and C may not be connected, but A is connected to B and B to C. What does this mean? In the case of parenthood, it means that A is C’s grandparent. If we follow the edges following their correct direction, we will discover all cases of direct ancestry. If we look to see whether two nodes are indirectly connected without bothering about the actual direction of the edges concerned, we discover all cases of blood relation. In the case of acquaintance, it means that A and C have a chain of mutual acquaintances, but they themselves need not know each other at all. A network is connected if any node in the network is reachable from any other, even if by a circuitous route. A hierarchy, such as a family tree, is a special type of network where all the edges have a direction, and there are no cycles, so the whole network has a direction, and we can consider nodes to be ‘above’ or ‘below’ others (see Figure 1.1).
A network is more valuable when there is a relatively high number of connections between nodes. If we have several such networks, they might also be connected to each other to create a bigger network (a network of subnetworks is itself a network). The connections between sub-networks might be relatively few in number (rather as there are lots of roads within cities and towns, but only one or two highways connecting them), but only one or two between pairs of networks are needed to create a connected network of networks. There is a branch of mathematics that quantifies network concepts, enabling us to work out when networks are more or less connected, which
A connected, undirected network
An unconnected network (or equivalently, two separate connected networks)
A directed network written as
A hierarchy
A network of networks
nodes in the network are more or less important in various different ways, but we will not pursue this type of analysis in this book.5
Metcalfe’s Law, named after Robert Metcalfe, a co-inventor of the Ethernet networking technology which helps carry the Internet, states that as the number of nodes in the network grows, the value of the network to the nodes in it grows exponentially (in practice, this is hard to sustain as the network gets very large, because it is unlikely that everything is connected to everything else, but the growth rate is still greater than linear).6 On the basis of Metcalfe’s Law, we should expect more people to get more value from the Internet as it expands, since each node on the network (a computer) is serving at least one person or organization. This disproportionate increase in value is called a network effect. Network effects also create a virtuous circle: as the network becomes more valuable, so more outsiders wish to join it. The network produces what are known as positive externalities; outsiders join in order to benefit themselves, but because of network effects they also benefit everyone else in the network without intending to.
Figure 1.1. Different types of networks.
POWER: NETWORKS VERSUS HIERARCHIES
Historian Niall Ferguson brings out the potency of networks in his book The Square and the Tower, using the image of a medieval city- state such as Siena, where the rulers sat in a tower exercising their power, and the people— merchants, artisans, peasants, artists, beggars—mingled in the main square below (open source software pioneer Eric Raymond makes a similar comparison between the industrial division of labour and the networks of the Internet, as the Cathedral and the Bazaar).7 The tower hosted a hierarchy, a top-down structure like a family tree; the square contained a network, a flatter structure with no restrictions on who is connected to whom. The rulers mainly interacted with their lieutenants, who passed on information about the conditions in the city, and who, given their orders, carried them out. While the arrangement has a certain parsimony, and for the rulers may be a rather nice deal, as a means of regulating the city it may be less than ideal. In particular, the amount of information received by the rulers is restricted by the number of lieutenants, and so their decisions may not be well-informed (especially as the lieutenants may themselves have interests that influence what they pass on). Put another way, knowledge of the state of the city can’t be gathered effectively by the relatively small number of people at the top of the hierarchy.8 Meanwhile, on the square, the rest of the city’s population mingled within a fruitful, innovative network structure. Artificial social distinctions may still count for something, but in principle anyone might bump into anyone else. Social problems, such as unemployment, crime, or widespread poverty, were evident to all. Information travelled quickly through the horizontal connections via gossip. Collective action happened through mutual support and reciprocity, not administrative fiat. And when the rulers failed to discern the mood in the square, and legislated inappropriately, a revolutionary situation might develop. Misunderstood patterns of behaviour in the networked square may then simply overwhelm the powers of the hierarchy to control them. This is not simply a medieval problem: the (hierarchical) French government has recently found it much easier to negotiate with the (hierarchical) trade unions than with the (networked) gilets jaunes 9
Hierarchies and centralization do have advantages. They provide a single point of control and order, so the city is more cohesive with a more defined identity and its government can move to a purpose (it is interesting that of Plato’s ideal cities in The Republic, the simpler versions are decentralized, but he has to introduce the hierarchy of Guardians and Auxiliaries to support both complexity and sustainability).10 The behaviour of a network is generally emergent it has properties which the nodes in the network do not have, and did not deliberately create—and so is hard to steer in a desirable direction, or away from the undesirable.11 Hierarchies are more easily planned, and can behave more strategically and consistently through time, following policies
through and pursuing goals (the COVID-19 pandemic spread through decentralized networks, and was battled by hierarchies). The ideal arrangement for many places no doubt is a hierarchy with a light touch, encouraging pluralism, markets, civil society, and spontaneous order, while ensuring that defensive measures such as healthcare systems, infrastructure, and military and police forces are maintained.
The balance, though, may be hard to achieve, and as the city grows, or becomes richer, or as the royal family tree produces more princelings who need a job, the constitutional arrangements may not scale. At that point, the square will take over, in effect by swamping the tower with information with which it will not be able to cope. The spreading of communications and trade will tend to build networks and leave the hierarchies behind, as has been argued happened in a globalizing world in the seventeenth century;12 hierarchies hit back through imperialism.
The Internet supports horizontal connections between networked individuals, and so gives them extra power relative to governmental and other hierarchies. This interplay between technology and social organization is a major socioeconomic change. ‘A steel mill could operate regardless of whether it was running under Nazism, under Stalinist socialism or under liberal democracy.’13 This is no longer the case, thanks to network technologies. Given the digital structures they have facilitated, the knowledge economy will function more smoothly under decentralized democratic structures than centralized ones. As one of the current authors argued elsewhere,14 this poses an authoritarian nation like China serious problems— the effective solution adopted by China, as we will discover later, was to co-opt and reshape the Internet in its own image, to the extent possible. This has meant changes to Chinese society, and relations between government and governed, as much as to Internet engineering itself.
FRAGMENTATION
Governments will struggle to regulate the Internet. When they desire, or wish to prevent, some kinds of outcome, they are tempted to legislate for (or against) those outcomes, in terms of incentives, prohibitions, enforcement, or targets. Yet there is no one to respond to the incentives on offer. The individuals in the network may not even be aware of the phenomena emerging from their individual behaviours, or may not feel responsible, just as drivers don’t feel responsible for traffic jams. A more intelligent type of engagement involves smarter regulation that influences the individuals rather than the network.15 However, this is not always an immediately more attractive approach for a government anxious to achieve a definite outcome, because it takes a lot of thought and isn’t guaranteed to work. Because the problem is emergent, it
may not be obvious how the individuals cause it;16 you may want to get rid of traffic jams, but equally you don’t want to stop everyone driving.
In particular, because it is a network of networks—i.e. sets of highly connected nodes with relatively sparse connections between them—badly designed legislation may make the Internet vulnerable to fragmentation, meaning that the connected network becomes unconnected. The fragmentation of the Internet is often discussed using terms such as the ‘Splinternet’ or the ‘Balkanization’ of the Internet. The virtuous circle that Metcalfe’s Law describes would become vicious, as users begin to drop off. Why would anyone break the Internet in two or more pieces? Governments might want to do it deliberately— some, like those of Iran and Russia, have toyed with the idea of having a national Internet (see Chapter 14). North Korea basically runs one, called the Kwangmyong.17
Suppose some scientists created an alternative version of TCP/IP that was incompatible with the old version— that might have the effect of Balkanizing the Internet. And this is not an empty worry— we will discuss later how certain nations and groups of nations have tried to alter the fundamental technological standards upon which the Internet sits. These basic standards have, so far, held, and this is a key factor in the Internet’s longevity and value. Part of the reason for this is that its positive network effects are so powerful.18
There has been much recent talk of the possibility of the Internet falling into Eastern and Western halves, the former run by China and the latter by the United States. Eric Schmidt, no less, the former CEO of Google, has predicted that this will happen.19 Different regulations would apply in the two halves, and—even if there were connections across the rift—it would still be complex, difficult, and slow to transfer data. Two completely separate information ecosystems might develop. There are some symptoms of this malaise appearing already; many Western sites are blocked in China, for instance, while Western suspicions of Huawei (see Chapter 12) could result in competing 5G standards.
In this book, we will explore the roots of this fear, which, unlike Schmidt, we believe to be unlikely. There are three reasons for this. Firstly, the situation is somewhat more complex than this— there are more than two significant views of how the Internet can be run. The world is not just America and China, although it may look that way from America and China. Secondly, the history and structure of the Internet shows that differing views can coexist simultaneously, and that there is room for different views, as long as we are prepared to accept diversity and cultural preferences. Pluralism is not out of reach on the Internet.
And thirdly, the notion of ‘fragmentation’ is not terribly well-defined. It looks relatively simple to fragment a network of networks. In Figure 1.1, the network of networks can be fragmented by removing one or two connections, but of course the Internet is far better connected than that. This means that,
