The BitTorrent Way?

Final Paper Participatory Culture

Gijs van Wiechen

Studentno.: 0000531

June 2004

Instructor: Shenja de Graaf

Introduction

File-sharing and P2P are a much debated phenomenon in the realm of politics, business, internet developers and cultural and communication studies alike. Whether it is a futuristic assessment promoting the cliché “Power to the People” or a fight against the illegal activities taking place, it cannot be denied that P2P has proven to be very influential in thinking about relations between humans and between humans and various aspect of our capitalist consumer society. The topic of this paper tries to deal with some of these implications, relating to a new file-sharing technology called BitTorrent. Introduced to the public in October 2002, it has grown considerably and has introduced a different way of sharing of content. One of the interesting things of BitTorrent is that its founder Bram Cohen never meant to facilitate such a vast amount of ‘illegal’ file sharing, and thus did not have to deal in its design with some of the problems that Napster and kin had and actively addressed. So while a part of the success is technologically embedded within the protocol, a large part deals with problems that are far more social, related to the behaviour and willingness of the users. The tension and interaction between the technological and the social, I will try to argue in this paper, shape to a large extent the different ways in which BitTorrent is used as a file-sharing tool and introduces interesting ways in which peers communicate and interact with each other. One of the most striking things of BitTorrent, for instance, is that users have to download a specific file in order to get access to content. One of the implications of this is that people are dependent on fixed servers that host these files, creating the possibility for interesting community-related interaction. We shall see that with the use of BitTorrent in various spaces of the internet, it can be seen as move towards what Clay Shirky has called ‘social sharing’. (Shirky, 2003)  Shirky sees this development as much in the download programs itself as in the use of it, making his term seemingly perfect for BitTorrent, although he does not mention it in his article.

Thus in this paper I will try to shed my light on what happens in the space between the original BitTorrent-protocol and the various user centred web spaces. I’ll try to examine in what ways the original protocol allows or obliges certain participatory acts, and in what ways these acts are shaped and addressed in various communities. First I’ll map out how BitTorrent works, and will search the ways in which BitTorrent introduces possibilities for new ways of (obligatory) interaction and hierarchies, and will position these developments within the debate on social aspects of file sharing. From this point of view I’ll try to pin point the implications of this protocol, by examining three specific websites, that all have a distinct way of managing access to files and ways of interaction between users. The first of these sites is Suprnova.org, one of the most popular sites related to BitTorrent, and its name has become almost synonymous with BitTorrent itself. I will then discuss Hawkies-world.org, which calls itself a BitTorrent-community. For this site membership is required and the content (of files as well as in conversation) gets actively moderated. Finally I will look at filelist.org a site that prides itself on being very restricted, not only moderating content, but actively monitoring members download behaviour and download/upload ratios. Furthermore access to their resources is restricted to members only and people who know to ‘work computers’. With these case studies it will hopefully become clear how a simple open-source program has introduced a way of file-sharing that reshapes current debates on social interaction in file-sharing and the ways in which P2P can function.

The BitTorrent Protocol and Original Client

Before we can start looking at the implications of BitTorrent and it’s relation to social practices in cyberspace, a short outline of what BitTorrent actually is, is of outmost importance. While BitTorrent has had a considerable growth since its public release in 2002, it remains an unknown phenomenon to many people, even those who regularly visit the internet or download files. Furthermore to grasp how BitTorrent obliges and promotes certain behaviour, it is very important to make clear what the original protocol allows. So below I’ll try to explain the intentions, practices and history of the BitTorrent protocol in short. I will not go into the technical specifics, unless this is of crucial importance to my argument.

Originally the BitTorrent protocol and client were released as a tool dealing with a problem very common with companies or people and their websites: server overload due to excessive traffic. Bram Cohen decided to create a protocol that dealt with this problem in a P2P manner. The sales-pitch on the official BitTorrent website sums this problem up quite well: “You have a great product, many customers, and are delivering your product to hordes of happy customers online. Serving large files creates problems of scaling, flash crowds, and reliability. As you grow, they become more central to your business, but your bandwidth costs go up as well. It's a vicious cycle.” [1] BitTorrent deals with this problem of server overload by tapping into the upload sources of the users downloading the file from a main server. The source that has got a 100% copy of a file available for downloading is called a seed in BitTorrent lingo. While a peer is downloading a file from a seed, he makes the already downloaded parts of the file available to other peers to download. Two simple figures make the benefits of this approach clear: The figure 1a shows the classic client-server model, and it is not hard to imagine the strain put on a central server when it is confronted with a large amount of clients. Figure 1b shows the ‘BitTorrent-Way’ wherein the pressure is divided between all the available parties. As is shown in figure 1b, a peer does not even have to be connected to the original seed to get the file.

figure 1 a: Client-Server                                                                   figure 1 b: BitTorrent

Source: www.bitconjurer.org/BitTorrent/introduction

To tap into such a file transaction, one must download a small file, with the extension .torrent. This file contains an outline of the number and size of all the different pieces of the main file (a so-called hash) and guides the BitTorrent client to a so-called tracker. This tracker is a web address that monitors the connections that exist around the transaction of a specific file. With this information the new user can make quick contact with some of the peers and seeders of the file. In the original protocol and client, uploading is the main condition for downloading. If a peer refuses to upload for one reason or another, it is impossible for him or her to download the file. This is an asymmetrical system, so the upload speed doesn’t necessarily have to be as high as the download rate. After a peer has downloaded the whole file, he or she changes into a seed and can choose to help the original seed distributing content. The word ‘choose’ is of significant importance, because while the protocol does force the peer to upload while downloading, it does not force the peer to leave his client running. The popularity of BitTorrent has mainly to do with the fast download speeds that can be reached, especially when one has a fast upload speed. This can happen because a peer is downloading from a multiple range of sources, but there is another important aspect. Without going into the specifics it is the efficiency of the protocol that tries to minimise the possibilities that peers have to wait in order to get a specific piece.  This is has among others to do with the process of ‘rarest first’, meaning that BitTorrent will give priority to relatively rare pieces, thus trying to create a balanced supply and demand for all the pieces.

This is a good time to posit a moment and look at the official BitTorrent client; probably the most used BitTorrent client out there at the moment.[2] First of all, the client is a very simple one, not unlike the normal download window of the Internet Explorer. It shows what and whereto it is downloading, the download process and speed and its estimated time left. Somewhat different than the IE window, is the mentioning of the upload rate and the amount uploaded. However as we have seen in the above, it leaves a lot of information out. There is among others, no mentioning of the amount of peers, the availability (according to number of seeds and their upload rate) and the address of the tracker. Other BitTorrent clients, using the open-source BitTorrent protocol, do show this and even let the user alter some of these aspects. I digress a little, but these differences are of crucial importance in showing clearly the intent that Cohen has. As he mentions in his resume, the fact that it is seamlessly integrated into the web is one of the important aspects.[3] To create a wide appeal (going beyond the file sharing scene) he has created a simple interface that is hassle free and fully integrated in Windows-dominated web surfing. It is clear that the BitTorrent client assumes a central seed and tracker that are on 24/7. When dealing with the far more decentralized use in the P2P-file-sharing scene, the information given is not enough. Because peers, seeds and trackers have a very irregular up and downtime in this realm, it is of importance for the peer to ‘know’ the persons and aspects involved. He’s got to know how many peers there are and whether there are any seeds left, or if the tracker is off- or online, in order to asses the odds of him getting this file. This again shows the commercial or at least mainstream intent Cohen has for BitTorrent and his version of a BitTorrent client. Not only the sales pitch and interface of the client prove this, but recently BitTorrent has made alliances with big companies such as Blizzard-software.[4] It is in this respect interesting to see that Cohen pitched the very first introduction of BitTorrent as an alternative to Napster, Gnutella and Freenet, promoting it on the famous Hacker-convention DefCon. Later declaring that “widespread copyright infringement was not what he had in mind”, when he made the client.[5] Whatever Cohen’s original intent however, the open-source aspect of the protocol and its enormous potential has given rise to different BitTorrent clients, such as Azureus, BitComet, BitTornado. These programs deal with BitTorrent in a more complex way, and let the user alter the specifics relating to a BitTorrent transaction. Having established how BitTorrent works and what Cohen’s intentions are, we can now start to formulate some basic assumptions.[6]

Flashmobs: BitTorrent in practice

It is important to keep in mind that a BitTorrent transaction consists of a swarm of seed(s) and peers around one specific file and one tracker.[7] The decision to publish a specific file is made by the original seed (also called the releaser) and it is relatively difficult for a peer to get exactly what he wants. To tap into a transaction a peer must first get a torrent file, and these files are most often located on a website of some sort. A BitTorrent user must thus scour sites looking for his content, because there is no one integrated search function that searches for the peer, as is the case with for instance KazaA and Edonkey. At this moment there is also no source (a website) that searches for the user through all the torrents available, and it is unlikely that there will ever be one, because of the high number and diversity of torrent-related sites. It takes thus more personal work from the peer to find its content with BitTorrent. We touch here on the decentralized aspect of this way of file sharing. Although KazaA and Edonkey among others do not have a list such as Napster had with ‘who has what’, we can say that all users do come together in one sphere. Because a search function is integrated in the program, users do not have to leave the realm of KazaA to get what they want. This program of course is installed and run from a personal computer, but still connects with a central network that connects different networks of peers. In more technical terms KazaA is still server, client and router integrated in one program.(Shirky 2003) This, however abstract it may seem, is one of the last centralized aspects of KazaA. The difference of BitTorrent maybe best illustrated with its relation to the Suprnova.org site. The BitTorrent client in itself is useless, to get content one most visit one of the numerous website and get a torrent file. It is thus not surprising that Suprnova.org has almost become synonymous with BitTorrent, simply because it is a main source of getting to content. Suprnova.org is, however, only a piece (be it a relatively big piece) of the proverbial iceberg. So while the KazaA network is decentralized in its structure which is then centralized on the user’s personal computer by the installed program KazaA, BitTorrent is only centralized in the user itself, making a transaction possible by always combining at least two elements from different sources. A user will always have to choose a BitTorrent client and a torrent file from a specific site.

At this point we might start thinking about the difference of BitTorrent in a commercial and a file-sharing context. The BitTorrent protocol is only a tool, and with this different uses it becomes obvious it does everything but dictate its use. As we have seen, Cohen coins his protocol as solving bandwidth issues, making systems and transactions more stable by sharing the burden with the clients. This is thus much a question of endurance, maintaining connections for long periods of heavy usage. When related to a corporate website and product for instance (say the Blizzard example), it thus generates a more stable situation that tries to limit overload and downtime. On the other hand, when BitTorrent is used in a file sharing context, a BitTorrent transaction is all but stable and fixed. Because of the relatively few persons involved in a transaction (certainly when comparing them for instance with KazaA or a corporate product), transactions have a tendency to become flash mobs. A (large) amount of people sprout up around a seed, file and a tracker, somewhere in cyberspace. They start downloading and seeding it, but when the initial rush is over, the number of peers will decline and so will the number of seeds. The original seed is often not in this transaction anymore, and has gone to seed something new. He is there mainly to set up a transaction by seeding it for a certain percentage (100+ of course), he can then choose to leave it to others. That this is common practice is visible on BitTorrent forums, messages from original releasers along the lines of “I’ll stop seeding when there are three seeds and then it’s up to you” are very common. The left over seeds often only seed a maximum of 100%.[8] The original seed was in essence doing an altruistic thing: uploading without downloading any new content. Follow up seeds will hopefully return the favour but it is unlikely that they share the extent of altruism, and abandon the transaction earlier then the original releaser. In time there will also be less demand and the transaction will die out. When the transaction is over, at first some peers will still tapping in, hoping on a new or reconnecting seed, but often these too will leave the realm, recognising that they have missed out on this file. What remains of the transaction is, frankly, nothing except maybe for a digitized reference on a tracker, but these also get cleaned up regularly: depending where it is published an old post may still linger at the bottom on a forum, leaving the remote possibility for a reseed. When the original releaser acknowledges this request the process starts up again, but this one will probably last even shorter. So in a specific time span a transaction will die out, leaving few references or traces. The time span of a working transaction can range from an hour to half a year or longer, depending on the number of peers and seeds, where it is published and how popular the specific file is. So while BitTorrent on the one hand allows a very stable and fast transaction, its durability is totally dependent from the different parties involved. If the main seed is server based, a transaction becomes equally fixed, if the seed is an individual, it will share his characteristics.

Seed/Peer & Content: Incentives to share

A lot has been written already on the paradoxical aspects on the social incentives of file-sharing, mainly focusing on ‘classic’ programs as Napster, Gnutella and KazaA.[9] Because of the emphasis put on this topic in the structure of the programs as well as in scientific research, it is important to see how we can relate BitTorrent to this discussion. The question that is central in the debate on incentives to share is two-fold. On the one hand there is the question why people are sharing in the first place. Sharing in this case refers directly to the issue of copyright infringement. Technically, and almost every sharer knows this, sharing copyrighted data is theft and punishable by law. It can lead to fines or even imprisonment. This is a very high cost for something not having a direct effect (meaning making content available does not necessarily supply you with fresh content). The other aspect is the problem of free-riding, or why people are downloading but refuse to upload. Different empirical researches have shown that, for instance, 25 and even up to 70% of the users of Gnutella do not share files. (Heyer, 2002; Adar & Huberman, 2002) While free-riding is a fundamental problem for P2P-networks, it is more important to look at why people are sharing and if these explanations make some aspects of BitTorrent more clear. The explanations can be narrowed down to three main aspects: an economical, altruistic and reciprocal incentive. Strahilevitz argues that the latter two are intertwined in a ‘code’ actively promoted by file-sharing programs but also lives in the mind of the user. (Strahilevitz, 2002) These are mutually enforcing, according to Strahilevitz: “the ingenious structure of file-haring networks solidifies a norm of sharing and that this norm of sharing is reinforced by the users’ mistaken but predictable notions of reciprocity”. Because sharing is of vital importance for a P2P network, Gnutella and others have created a ‘charismatic code’ promoting the fact that everybody shares. This as empirical data shows is a plain and simple lie, but by stating this lie, it taps into the user’s feelings of reciprocity. If you have been given something by someone, supposedly as an altruistic deed, it is more then kind to give something back. This code however also masks uncooperative behaviour, by only showing the persons who do share to an individual user. Furthermore it creates affinities between large groups of anonymous users, simply because people are meeting each other because they have the same interest. Strahilevitz acknowledges however that this works more strongly when dealing with relatively obscure artists, but it remains an important aspect by which in the minds of the user a picture is created. In some file-sharing applications (for instance KazaA) this is even more advanced, and users can search other user’s lists, to try to find something interesting. Not dealing with these social factors, Becker & Clement argue that there is also an economical motive. This has to do with the rational process of keeping the network alive. This incentive is particularly strong when a network has just started, because the amount of the same file and the diversity file are responsible for the success of the network. Becker & Clement also see a difference between the creator of a file, and someone who’s merely passing it through to other persons. While the ‘cost’ or work of the first is relatively high, the latter will have to do profoundly less to share it. This analysis however seems to fall short, because again it seems that social considerations dictate the willingness to help a network grow.

BitTorrent deals with these incentives to share, but we have already seen that it can not be considered as one network. One crucial difference moreover is that with BitTorrent uploading is obligatory for downloading. So while KazaA and Gnutella deal with the fact that some persons just never share, because they have nothing to upload in their shared folder directory, with  a BitTorrent transaction a peer will always share something, ranging from 0,1 % to over a 100. The incentive is thus mainly focused on the fact that a peer must leave his download window open after he finished downloading. It seems to be that an original releaser, even more than with KazaA or Gnutella, makes a decision to put effort into a release solely on altruistic grounds. This altruistic behaviour is enforced by the high visibility the first seed has and the fact that he can actively monitor the number of peers. When publishing on a large file-sharing network, the file quickly becomes a part of this network, and the publisher has technically lost touch with his product. Although this can happen with BitTorrent as well, and it often does, this is only after some time, with a structural break in between. As we have seen a BitTorrent transaction starts every time from scratch, and as mentioned above a lot of seeds are actually choosing to stop seeding when the transaction can continue without him. With an over the top comparison, it could be seen as feeding a baby bird, only to watch it fly away, giving the nice feeling that you did a good job. Of even graver importance, however, is the fact that a lot of transactions are started within communities that already established a relatively dense-knit group. Because of the large amount of forums and communities on the internet, a user can choose and join one or more that has his or her taste(s). There are, for instance, BitTorrent-forums related to language, ethnicity or nationality, forums that are specialised in a specific movie or music genre or dealing with one specific program.[10] While reciprocity on Gnutella is for Strahilevitz largely based on misperception, and anonymity a force to be dealt with, a BitTorrent-forum actually shows a relatively large percentage of people you are dealing with. This is not to say these people actually know each other, they are still largely nicknames sending over-friendly Thanks or comments, but it becomes clear that other user are similar people as the user. When you know that a member called “Megaman” has supplied a release you enjoyed, the best way to thank him is to share some of your own interests and the forums give you the possibility to actively recommend it.  The way which interaction is structured on a site or forum is, however, is very diverse as we shall see in the different case studies.

This rather positive and optimistic analysis goes by the fact that BitTorrent has a free-rider problem too, and it seems that much in the same way as Becker & Clement describe it, this problem gets larger as the BitTorrent community grows. In BitTorrent lingo these persons are often referred to as ‘leechers’. Because the BitTorrent protocol does not dictate an upload/download ratio and only forces to upload until the transaction is completed, a charismatic code such as Strahilevitz argues is not a universal phenomenon. People are dealing with this problem in a far more diverse and decentralised way. Rules and promotion related to the encouragement to remain in the transaction after their download has been finished, happens on a variety of different levels. In the popular alternative BitTorrent client Azureus a warning is integrated through a pop-up, announcing that the user’s upload ratio for a specific file is less then 100 % and that a ratio of under 100% is a bad thing for the BitTorrent network. More often the website or members of communities promote seeding, or are portraying leechers rather negative.[11] The case studies will dedicate some attention to these aspects in the three different websites.

Narrow Horizons & Darknets: Trust and Accountability

Clay Shirky and Robert Kaye have both written about the future of file-sharing and the phenomenon of Darknets. (Shirky, 2003; Kaye, 2004) They bring forward some interesting problems that we can relate to different BitTorrent transactions and are very useful when we start looking at the specific cases. Shirky argues that with the new tactics of the RIAA (Recording Industry Association of America) trying to ‘crush connectors’ such as KazaA, file sharers will move into tighter groups that are difficult to penetrate. Because KazaA is maintained by a relatively small number of highly connected nodes, if the RIAA tries to attack these, it brings much harm to the system. When this tactic succeeds, and according to Shirky it does, it’ll force the file-sharers to go somewhere else. Because the threat of attacks is becoming larger when a network is highly visible, Shirky envisions that people move into highly secluded groups unobservable from the outside world. This move of file-sharers into ‘underground groups’ is also signalled by Robert Kaye. A term often used for this mixture of file-sharing and social groups is a so called Darknet. According to Kaye “combining file sharing applications with social networks enables people to create a trusted network of their friends to keep out he bad guys”. Trust and accountability have been major issues in P2P since its beginning. With the RIAA and others trying to infiltrate large networks, the trust of the peer-to-peer transaction is under great strain. Not only can a file be incriminated a seemingly harmless peer can log your IP and use it for a lawsuit of some sort. Trust is difficult topic but two aspects are of crucial importance. One is the fact of reputation; the other has to do with the number of people involved. (Waldman & Cranor, 2001) If a positive reputation of a releaser or site is established, it is easier for the user to trust the persons involved. The second factor is the number of person involved; it speaks for itself that fewer involvements of unknown people will reduce the risk of one being a ‘bad guy’. With their actions the RIAA thus forces file-sharers to find a different way of interacting. Accountability is a second important aspect, and more or less relates to the problem of trust: who is responsible for the content and transaction. (Dingledine, Freedman & Molnar, 2001) This also has to do with reputation; a familiar person will be more likely to give you good content. Content in large networks are often unaccounted for, and mislabelled content seems to be far more common today than old research suggest. [12] The RIAA and music labels have been releasing corrupt copies of popular songs on KazaA and other instances (on all the large networks) and there are many examples of files that do not consist of the things promised. Secondly the availability remains uncertain until a download is finished on these networks.[13]

Shirky and Kaye are quick to acknowledge the problems of turning to small and secluded groups of file-sharers generates. One the one hand fewer members in a community, obviously means fewer files available and could endanger the popularity of this system. Furthermore fewer members could seriously harm the efficiency of a system. But on the other hand it allows users to explore their weak and strong ties, where in older, massive file-sharing systems there were only ‘random ties’. The benefit of this is, according to Kaye “Exploring the weak ties in your network is likely to give you access to more relevant information/content than a random tie”. This allows people to show interest in one another and their tastes, more than solely searching for personal benefit.

BitTorrent is not mentioned in Clay’s article, and Kaye sees it as being one of the tools that the ‘darknets’ could use. Again we encounter the problem that while BitTorrent is extremely useful for these tight-social networks, these aspects are not inherent in the protocol itself. It’s its use that dictates the way how it works. For instance, while Kaye and Shirky are emphasising the protective aspects of these new groups, BitTorrent itself is a highly unprotected way of sharing files, with every IP in the transaction easily traceable, and the original seed(s) for all to see. To deal with this security issue, users and communities have to look for other ways to protect their actions or remain secluded. Firewall programs like protowall or peerguardian must be installed if a member wants to be safe from bad guys peeking in, communities can do essentially the same thing by monitoring IP and blocking access to possible bad guys. However because the transaction is so focused on one file, and it can reach fast download speeds, it wins at this point in accountability. The fact that torrents are published by community members introduces furthermore forms of reputation and trust among members, especially when the communication is taking place in a members-only community. Looking at these different forms of communities surrounding BitTorrent, it becomes clear that this realm is still constantly developing and recently there is a shift to more secluded realms. Many web communities have forbidden their content to be published on highly visible sites such as Suprnova.org, with members that get caught risking a permanent ban. Again it will be the case studies that will define these characteristics more clearly.