Peer-to-Peer : Harnessing the Power of Disruptive Technologies (Hardcover)

Five years ago, I proposed a filestore that would be impossible to censor because it was too widely distributed across the Internet. I called this `The Eternity Service' after the `Eternity circuits' described by Arthur C. Clarke in `The City and the Stars'. I had been alarmed by the Scientologists' success at closing down the Penet remailer in Finland; this showed that electronic publishing can make it easy for rich people with ruthless lawyers to suppress publications.

Gutenberg's invention of print publishing made it impossible for princes and bishops to censor troublesome books, but might electronic publishing not make it possible once more? If there are only half a dozen servers containing a controversial document, then court orders can be purchased to close them down. Might not electronic publishing compromise the enormous gift that he gave mankind?

I could never have imagined the effect my paper would have. Rather than sedition, blasphemy or pornography, the battle is being fought over music copyright. Thanks to the Recording Industry Association of America, and its lawsuit against Napster, the Eternity Service has spawned a host of peer-to-peer systems such as Gnutella, Mojonation and Publius that have become front page news. There are also less-well-known systems, such as Red Rover, whose goal is to enable dissidents in places like China remain in contact with each other and with the rest of us. Sometimes, I have felt a bit like a skier who sets off an avalanche, and can only watch in fascination as it thunders down the valley.

So I eagerly awaited my advance copy of this book, and I have certainly not been disappointed. Although it has been put together rapidly by a number of different authors, this is not just a list of what systems X, Y and Z do, and how they work.

The first chapters place the peer-to-peer movement in a broader context: the early Internet was peer-to-peer, as was usenet. More recently, IP address congestion has led the builders of systems such as ICQ to design their own namespaces; alternatives to DNS are one of the emerging features of peer-to-peer.

After the introduction, there are chapters on a number of different systems: SETI@home, Jabber, Mixmaster, Gnutella, Freenet, Red Rover, Publius and Free Haven. Although many of these systems are relatively young and evolving quickly, their successes and failures to date help sketch a map of the peer-ro-peer space as of the beginning of 2001. They teach us what has worked, what hasn't, and what just might.

The lessons learned are distilled in the last seven chapters of the book. The themes brought out here have mostly to do with how these services can be made predictably dependable. Performance and trust are intricately intertwined; only by having means to protect against a wide variety of flooding and other attacks can service be maintained in the face of hostile action. However, technical mechanisms alone are not enough; there will probably have to be economic and social incentives for users to behave in apropriate ways. This is very timely, as people have recently come to recognise the major role played by perverse economic incentives in creating and maintaining information security vulnerabilities.

I believe that this book will become essential reading for a surprisingly broad range of people. Peer-to-peer issues do not merely affect designers of systems such as freenet, but potentially any system that has to deal with large scale and intermittent connectivity. If you believe that within ten years we will see all sorts of devices from burglar alarms to fridges and heart monitors able to establish ad-hoc communication with each other using mechanisms such as Bluetooth, then the next thing to consider is how these ad-hoc networks can be defined and protected. Here, the lessons from peer-to-peer may be invaluable.

If you need to understand the peer-to-peer movement - where it came from, where it's going, and what it can teach those of us who are working in related fields - then this book is a must-read. I also think it deserves a place on the shelves of everyone doing serious research in computer science.

Ross Anderson, Cambridge University Computer Laboratory

When I first picked up the book, I had modest expectations. I have been disappointed by technical experts treating topics from the social sciences -- and this book does just that. Different chapters focus on such issues as: incentives on users to cooperate, the vulnerability of computer networks to social control, strategies for reliable communications, and censorship. Yet in this volume each topic is treated clearly, intelligently, and insightfully.

The authors not only summarize their topics well, they regularly offer sparkling insights. For example, in the chapter "The Cornucopia of the Commons," Dan Bricklin explains how certain peer-to-peer applications are enriched by consumption. The more that users consume from the electronic commons, the larger that electronic commons becomes. In the case of Napster, as users download files those files become part of the overall archive available to others. This turns the tragedy of the commons on its head: well-designed peer-to-peer applications can create explosive processes of value generation - an insight I find both provocative and profound.

The book sits squarely at that most difficult spot on the intellectual spectrum: the place where technology and policy overlap. Is this a policy book? Yes, it is. The topics above are all policy-relevant, and for a technical expert many of them would be new. Is this a technology book? Yes, it is that, too. It talks about network architecture design, technical implementations of trust and reputation, name spaces, and searching. For social scientists, the book is an excellent introduction to computer networking.

Peer-to-Peer is nearly 400 pages long and has 19 chapters. Amazingly, every chapter is worth reading. I can't say that about many edited volumes that I know! The editor also did a good job of integrating the different chapters so that the book has overall coherence.

This book is perfect for a university-level class about the Internet. The chapters on name spaces are useful to study of ICANN and global governance. Chapters on Napster help when studying intellectual property, those on FreeNet are useful when studying free speech. In my Internet policy class, I sprinkle chapters from the book throughout the semester.

Aside from teaching, the book is useful for anyone who wants to understand computer networking. It is accessible and readable, yet surveys a wide range of technical topics.

Considering the importance of the Internet and of peer-to-peer networks, it can be surprising difficult to find good explanations of the issues. This book does just that.

###

On September 2000 O'Reilly organized a peer-to-peer summit with a number of experts (computer scientists from MIT and AT&T Labs, CTO's, architects, human rights activists). This book is basically an offspring of this summit, with contributions from many of those experts.

One of the goals of this summit was to answer the question what peer-to-peer really is, and about what technologies we should think when we hear the term. There are also a lot of lessons to be learned from well known applications as Napster, Gnutella and Freenet. One of the outcomes is peer-to-peer is much more than file sharing; we also can think of projects regarding Web Services (Bluetooth, .NET, JINI), instant messaging, Web hyperlinking and networked devices. Core benefits include "more effective use of Internet resources through edge services" and "overcoming barriers to formation of ad-hoc communities and working groups". Peer-to-peer should be positioned as a natural next step in the development of the Internet.

After setting the context the book continues with chapters about a number of systems (already developed or still on the drawing table): SETI@home , Jabber, Mixmaster Remailers, Gnutella, Freenet, Red Rover, Publius and Free Haven. The focus of these chapters is on high level requirements and design choices: what works, what does not work and why?

The last part of the book contains descriptions of core technologies and research areas. These chapters deal with questions as performance and scalability, search strategies, accountability, trust, reputation and security. The approach of these topics is scientific: performances issues in Gnutella and Freenet are illustrated mathematically via graph theory. In the chapter about accountability several techniques for digital payments are described for ensuring accountable behavior amongst peers. These chapters also show that peer-to-peer is not an isolated area, many of the requirements also apply to other Internet related areas: think about encrypting e-mail or authentication of users by a server ("How do I make sure that only persons above 21 years old can look at my Website" or "How can I trust the person I am dealing with at an auction site").

This book is clearly not meant for developers. This is a book for people who want to be up to date about developments in computer science. Hereby I am thinking about CTOs, architects and researchers. In my opinion this is a groundbreaking work about a new and important topic that has many people's attention. The level is high and academic, and many of the technologies actually would require a separate book. All the articles are well written and of good quality.

I want to place one critical remark however. One of Tim O'Reilly's goals was to show the world that peer-to-peer is much more than violation of copyrights, or other activities that might be viewed as subversive in one way. However, in this book applications for file-sharing and censorship resistant publishing are overrepresented. There are many categories of worthy projects not included, such as "Servers/Services as Peers", "Devices as Peers" and "Writable Web". In order to get peer-to-peer on the map as more than a fringe technology , we also need attention for applications that are more garden variety, affecting our daily live or useful within a business context. Hereby I am also thinking about UDDI, JINI, Bluetooth, .NET and WebDAV.