An empirical study of Namecoin and lessons for decentralized namespace design Harry Kalodner*, Miles Carlsten*, Paul Ellenbogen*, Joseph Bonneau†, Arvind Narayanan* {kalodner, carlsten, pe5, arvindn} @cs.princeton.edu* [email protected]† Princeton University Abstract 1 Introduction Decentralized namespaces. We initiate the study of decentralized namespaces from an eco- Secure decentralized namespaces have recently nomic perspective. A namespace, as we define become possible due to cryptocurrency technol- it, is an online system that maps names to val- ogy. They enable a censorship-resistant domain- ues. The Domain Name System (DNS) is the name system outside the control of any single most prominent example. A web service such entity, among other applications. Namecoin, a as Twitter that allows users to claim usernames fork of Bitcoin, is the most prominent example. and create profiles can be thought of as imple- menting a namespace. To be memorable by hu- We initiate the study of decentralized names- mans, namespaces must support arbitrary user- paces and the market for names in such sys- chosen strings as names. To be secure, names- tems. Our extensive empirical analysis of paces must map each name to the same value Namecoin reveals a system in disrepair. Indeed, for all users and adversaries shouldn’t be able to our methodology for detecting “squatted” and convince a user that any other value is correct. otherwise inactive domains reveals that among The problem of decentralized namespaces has Namecoin’s roughly 120,000 registered domain long been recognized as an important one. The names, a mere 28 are not squatted and have non- DNS is a critical yet centralized component of trivial content. Further, we develop techniques the Internet and those who control it can alter for detecting transfers of domains in the Name- the web for all users. The controversies around coin block chain and provide evidence that the the shutdowns of wikileaks.org and the market for domains is thin-to-nonexistent. 2011 domain name seizures by the U.S. DoJ and DHS illustrate why many researchers and We argue that the state of the art in mecha- activists have sought decentralized alternatives nism design for decentralized namespace mar- [1]. kets is lacking. We propose a model of utility The above three properties — security, user- of different names to different participants, and chosen names, and decentralization — are articulate desiderata of a decentralized names- known as Zooko’s triangle [2]. Until 2011, de- pace in terms of this utility function. We use signing a system that exhibited all three was this model to explore the design space of mech- conjectured to be impossible [3]. The rationale anisms and analyze the trade-offs. was that enforcing the uniqueness of name-value 1 mappings and a consistent view of the directory and what are the goals of mechanism design for for all participants would require a centralized namespaces? How well does Namecoin succeed server or a hierarchy. at attaining these goals and what are its limita- Cryptocurrencies and Namecoin. Cryp- tions? If Namecoin is not the ideal design, can tocurrency technology enables building a we analyze the design space as a guide to the namespace with all three properties. Put simply, creators of future decentralized namespaces? the block chain is a global, distributed data Our contributions. Along the above lines, structure that can be repurposed as a directory. we make the following contributions. We be- Miners execute a consensus protocol to estab- gin by proposing a model of utility of differ- lish the state of the system and are incentivized ent names to different participants and articu- to do so by mining rewards they receive in lating desiderata of a decentralized namespace exchange for their participation. As long as a in terms of this utility function (Section 3). majority of miners — weighted by computing We highlight the difficulty of mechanism design power — follow the protocol, all users will see even in a toy model and explain the importance a consistent view of the directory when they of making the model more realistic by incorpo- query it. This in turn gives the system, and rating extensions such as time-varying prefer- hence the underlying currency, economic value, ences. making miners’ actions profitable to them. The central contribution of this paper is a Namecoin is a cryptocurrency that realizes a thorough empirical analysis of the Namecoin decentralized namespace. It is the first altcoin ecosystem. We develop a series of criteria based from Bitcoin with its own block chain. It offers on the block chain, network behavior, as well the same features as Bitcoin with the addition of as content to distinguish active websites from a name/value store that can be used to hold arbi- parked or squatted names (Section 4). This al- trary data (see Section 2). The name/value store lows us to iteratively filter our dataset of around supports various applications; primarily, Name- 120,000 registered names in Namecoin, leaving coin has been used for domain-name resolution a mere 28 that are not squatted, and have non- for the ‘.bit’ alternative TLD, and by the online trivial content. identity service, OneName, which utilizes the We then delve deeper into the economics of Namecoin block chain to record data about its names. Namecoin has a built-in ability to trans- members. fer names, which is a secure way to trade them Namecoin offers a novel solution to the using the block chain. However, it is not obvi- technical challenges of decentralized names- ous which transactions correspond to such sales, paces. However, there are also economic chal- as opposed to regular name updates. We develop lenges. These arise from the fact that even a novel analytic technique to distinguish the two though namespaces theoretically support an in- types of transactions and find evidence for about finite number of names, the supply of names 250 transactions in the entire history of Name- that are memorable and meaningful to humans is coin that may represent sales of names (Section scarce. Allocating these names to users is there- 5). Of course, it is possible that more sales have fore a mechanism-design challenge. The central occurred off the block chain, but there doesn’t thesis of this work is that this mechanism design appear to be a widely known marketplace for challenge is far harder than realized. A system Namecoin names. that gets it wrong may “work" in a narrow tech- Based on all the empirical evidence we nical sense, but may not be useful to real users. present, we are left to conclude that the Name- Specifically, there are several crucial ques- coin ecosystem is dysfunctional. The vast ma- tions to consider: how do we model the eco- jority of registered names represent squatting nomic behavior of the users of a namespace and there is little evidence of a secondary mar- 2 ket for names. While there could be many fac- and server software, users, a mechanism, and so tors that explain the lack of adoption, there ap- on. In fact, Namecoin contains namespaces in pears to be clear room for improvements in the the computer science sense, which we term sub- design to minimize squatting and other prob- spaces in this paper. lems. To this end, in Section 6, we explore the design space of decentralized namespaces and make recommendations. 2.1 History Why study namespaces? Although we find Namecoin is an alternative cryptocurrency, or that the Namecoin ecosystem is in disrepair, altcoin, modeled after Bitcoin [4]. Furthermore, studying namespaces is important. While there it is the first altcoin in the sense that it was currently doesn’t seem to be widespread dissat- the first to create its own block chain, sepa- isfaction with today’s DNS causing users to seek rate from Bitcoin’s. Namecoin shares many censorship-resistant alternatives, the existence similarities with Bitcoin, including the same of such alternatives provides a valuable hedge method for proof-of-work, the same coin cap, against a potentially abusive central authority. the same block creation time, and all of the same Besides, domain names are just one application transaction operations (with a few additions). of namespaces. Centralized directories for user Namecoin was inspired after discussions about public keys have fared much less well than DNS, a BitDNS [5] protocol using a block chain to and the service OneName, which we discuss manage a domain name lookup service. The in Section 8, is an interesting alternative. And motivation was that a central authority manag- of course, the problems posed by namespaces ing domain names, such as ICANN, requires too are intellectually interesting to study. We think much trust in a single entity and represents a sin- decentralized namespaces have many important gle point of failure. The first Namecoin block applications, but their promise hasn’t been real- was mined in April 2011, and as of this writing, ized so far. Our work helps understand why this over 215,000 total blocks have been mined in might be and lays the groundwork for a more the Namecoin system. Because of its similarities rigorous approach to building such systems. with Bitcoin, Namecoin was able to be merge- mined and has been merge-mined with Bitcoin since October 8, 2011 (see the next subsection). 2 Background: Namecoin In this section we cover the background of 2.2 Description of the block chain Namecoin. We begin by looking at the history of Namecoin is minted and maintained by a decen- the cryptocurrency and then explain why a block tralized peer-to-peer network.1 Namecoin trans- chain can be used for our definition of a names- actions require the digital signature of the ac- pace.