The synthesis of Lave & Wenger’s (1991) “communities of practice” with the Chicago School writing about work and practice presented in chapter 1.1 provides a good model for thinking of how a practice is reproduced in time.149 This model, however, sheds little light on reproduction of practice in space. In chapter 1.2 we looked at the work of software developers in Rio de Janeiro, noting complex interactions that cannot be understood without considering the fact that those developers are engaged in a practice based far away from where they live and work. Furthermore, considering their peripherality relative to the practice is not only important for understanding how their work fits in the larger world of software development. Rather, we cannot understand what happens locally, in Rio de Janeiro, without considering this peripherality.

      Before turning to the way peripherality structures local interactions, however, I ask a few broader questions about reproduction of practice across space. How does a practice spread in space, sometimes extending its membership to people who live thousands of miles away? How does software development, a practice once confined to the east coast of the United States,150 come to be accepted as something sufficiently universal that a youth in Rio de Janeiro can aspire to be a software developer without making people laugh?151 I then look at the local politics involved in such reproduction.

      Approaching the question of how practice spreads across space from the perspective of a Chicago School reading of Lave & Wenger (1991), we might think of practice as spread and synchronized by itinerant practitioners. Indeed, the role of such people has been described by some authors. Saxenian (2006) talks about the “new argonauts” who build links between the silicon chip industries of California and Taiwan, or between the software industries of California and India. Similar travel of practitioners is described in some of the studies of physicists (Traweek 1988/1992, Knorr Cetina 1999). While looking at traveling practitioners is increasingly important for understanding the sharing of certain kinds of practice, such people are still a fraction of world’s population and their travels tend to link rather specific location (for instance, San Francisco and Bangalore). No matter how close San Francisco may be to Bangalore, most cities in the world, even large ones like Rio de Janeiro, are far from both in the experience of most people who live and work there. At least in the case of the software developers working in Rio de Janeiro, it is clear that most of the work of keeping the local practice “up to date” is done by people who rarely (if ever) leave Brazil — the case of most people whom we met in chapter 1.2.152,153 And while the traveling practitioners may well be crucial for the initial introduction of the practice in a new place, I will try to show that we cannot easily separate this initial step from the later synchronization. Reproduction of a practice involves setting up a system of relationships between people and objects. It cannot be achieved by simple arrival of individual practitioners.154

      The concept of “networks of practice” (Brown & Duguid 2000, Duguid 2005) aims to extend the notion of “communities of practice” to groups of practitioners scattered in space, and allows for electronic communication between the practitioners. As Duguid (2005) points out, however, successful communication in such networks assumes prior commonality of practice.

The network of practice (NoP) designates the collective of all practitioners of a particular practice. For example, Knorr Cetina’s (1999) “epistemic culture” of high-energy physicists constitutes a global NoP that has within it multiple local CoPs [communities of practice]. Though practice is not coordinated within a NoP as it is in a CoP, common practices and common tools allow distant members to exchange global know that and to re-embed it (Giddens, 1990) in effective, coherent ways through the mediation of their locally acquired knowing how. Consequently, where practice precedes it, explicit knowledge may appear to have global reach (or to be “leaky”). Where it does not, the same knowledge may appear remarkably parochial (or to be “sticky”).

The central distinction between the CoP and the NoP turns on the control and coordination of the reproduction of a group and its practice. Newcomers enter the network through a local community. You become an economist by entering an economics department in Chicago, or Berkeley, or Columbia—a route that may mark you for life, in part because the tacit knowledge of the local community profoundly shapes your identity and its trajectory. (p. 113.)

In other words, where shared practice exists, the practitioners can share knowledge and further synchronize their practice through a range of means, including electronic communication. Someone living in Brazil who is experienced in the practice of software development as done in California (that is, fluent in the culture and language of that practice and comfortable with its tools) will have no difficulty understanding documents authored by software developers in California. (Though, they may still have some trouble in applying this understanding to the local context.) But how do they become experienced in that practice in the first place and why do they even want to engage in it? Duguid’s answer (that they enter the practice though a local community) points us in the right direction. It assumes, however, the prior existence in different places of local communities that can be said to engage in the same practice. We need to understand how this is possible and what it even means for the “same” practice to be carried out in a new place, looking both at the way those communities achieve substantial similarities of practice and also at how the claims to participation in the same practice are asserted and judged. Duguid (2005) writes, for example, that one becomes an economist by entering a department at Chicago, Berkeley or Columbia. We may want to ask, however, whether one can become “an economist” in the same sense by entering Instituto de Economia at the Federal University of Rio de Janeiro or the Faculty of Economics of Eduardo Mondlane University in Mozambique. In other words, the question of how one becomes an economist cannot be treated separately from the question of what counts as “being an economist” and who gets to do the counting.

Abstract Groups of Practitioners

      To answer such questions we first of all must recognize the crucial analytic move introduced by Duguid (2005): the quote above drives a wedge between the notions of “practice” and “local community.” To push this wedge further we need to look more carefully at the concepts of “groups” and “communities” invoked in the previous chapter, and the different levels of analysis implied in the Chicago school view, in particular distinguishing between concrete communities (typically tightly knit and local) from abstract groups (or “categories”) of people united by similarity of practice — not forgetting the all-important question of who acts as the judge of similarity. To draw on Becker’s (1963) example, we can look at the individual dance musicians as members in a concrete group, such as “the X— Avenue Boys” (whom Becker describes as “a clique of extreme jazzmen”), or in the more abstract groups such as “the dance musicians.” Once this distinction is drawn clearly, we can examine the relationship between the two. How does being one of “the X— Avenue Boys” make one “a jazzman”? Conversely, how does being “a jazzman” allow entry into specific local cliques?

      To illustrate, the software developers I interviewed in Brazil, often asked me if I was a software developer myself. When they did this, they were not inquiring whether I was a member of some local “occupational community” or “community of practice” — they usually knew that I was not.155 Instead, they wanted to know whether I belonged to a larger, more abstract category of people who write software code and with which they themselves associated. As a foreign member of this group, I was not expected to understand local meanings and norms. For instance, the developers took time to explain to me the many difficulties of doing this kind of work in Brazil. They also pointed out specific people in specific organizations that I should talk to — again, correctly assuming that I would not know by myself who the important people were. At the same time, they expected me to understand their technical jargon (when used in English), as well as certain values and practices. For example, having identified myself as a “former software developer,” I was expected to not just know what a “source control system” is, but also the technical and the social implications of the statement that a particular company lacked one. (At the technical level I would need to imagine the likely outcomes, while at the social level I would be expected to form the appropriate opinion of the people who run the company.) In fact, I often needed to make special effort to make my Brazilian interviewees suspend the assumption that I share their meanings and opinions and to explain everything to me, as if I was not one of them.

      This willingness to assume that I would understand their terms and practices is not a matter of wishful thinking. The practices of Brazilian software developers do replicate those of software developers in other countries to a remarkable degree. As I will show in more detail later, replicated practice involves not just solving the same problems, but solving them in the same way, relying on the same set of concepts, calling relevant objects by the same names (either in English or by Portuguese terms borrowed from English), and making many of the same jokes in the process. In an important sense, Brazilian and American software developers are members of a large collective of people, spread around the world, who write software code, and we must recognize the similarity of practice within this group must be recognized. At the same time, we cannot take for granted either this similarity as observed by the ethnographer, or the fact that Brazilian developers’ claims to similarity of practice are accepted (usually) by local outsiders and remote colleagues.

      So far in this section I have used the terms “abstract groups” and “categories” interchangeably when referring to larger classes of people engaged in the same practice. I do so to highlight the dual nature of such collectives: they are “groups” that exist objectively, identified by similarity of practice and communication, but they are also “categories” in the sense that they are named groupings that are recognized by social actors. This distinction corresponds to the one that Lamont & Molnár (2002) draw between “social boundaries” and “symbolic boundaries.” Symbolic boundaries, are “conceptual distinctions made by social actors to categorize objects, people, practices, and even time and space” (p. 168). Such boundaries correspond to what I call “categories.” Social boundaries are “objectified forms of social differences,” which are “revealed in stable behavioral patterns of association, as manifested in connubiality and commensality” (p. 168). Social boundaries define what I call “groups.” Lamont & Molnár argue that symbolic boundaries are crucial for continued existence of social boundaries.156 (This argument is similar to Giddens’s theory of structuration discussed below.) Both types of boundaries are “real.” Similarly, classes of practitioners are simultaneously “groups” and “categories.” This duality is part of the reason why I avoid Brown & Duguid’s (2000) term “networks of practice” — the concept of “networks” in sociology typically focuses on the patterns of interaction ignoring the symbolic boundaries perceived by social actors.

      Abstract groups of practitioners are therefore real not only in the sense that they represent actual similarity of practice, but also in the sense that successful claims to membership in such groups (or categories) have rather tangible effects on one’s interaction with other members as well as with outsiders — perhaps even more so with outsiders who are less likely to recognize the local cliques. This effect is distinct from that of membership in concrete local group. In some situations, perhaps when interacting with the closer associates, what matters is membership in the small group. (Chapter 3.1, “Aplicações Corporativas,” starts with a vignette that shows how membership in a small clique, “the Herculoids,” opens for me the doors of “Alta.”) In other cases, for instance when looking for a job (or simply presenting oneself) outside one’s immediate social group, it may be more important to be “a policeman,” “a jazz musician,” “a fisherman” or “a developer.”

      While some of the different authors that we considered in chapter 1.1 stressed the more concrete groups and other stressed the more abstract ones, they rarely draw clearly the distinction between those two types of groups. Hughes’s discussion seemingly focuses on abstract groups (“the doctors”). Becker’s Outsiders (1963) also mostly deals in abstract terms (“marijuana users,” “jazz musicians”) and while concrete “cliques” are mentioned and sometimes described in some details (“the X— Avenue Boys”) the relationship between the clique and the larger abstract group is left unexamined. The early discussion of communities of practice (Lave & Wenger 1991) similarly leaves the level of analysis unclear. The book often refers to the larger groups, operating at a regional or national level (“Yucatan midwifes,” “Liberian tailors”, “butchers”), yet the mechanisms that it proposes often index individual interactions and would be more appropriate if the group were to be understood as small and local. Legitimate peripheral participation goes a long way towards explaining how novice comes to practice tailoring in a particular small community of tailors, but it leaves us wondering how each of them becomes “a Liberian tailor.” Orr (1996) focuses explicitly on the smaller local communities but mostly avoids the discussion the any larger groups that “the reps” are parts of.158

      Van Maanen & Barley (1984) come closest to recognizing the need for clarity on this issue. Dissatisfied with the discussions in terms of abstract categories, they call upon researchers to look at the smaller (and typically local) groups. Their examples include the different subgroups within the “traditional fishermen” of Gloucester, Massachusetts (e.g., “Guineas” and “Greasers,” each with its own occupational culture), “sweepers who live together in closed communities in Benares” (p. 303) and “cab drivers in Boston” (p. 311) — groups often defined by place. While this attention to links between occupational groups and place is important, Van Maanen & Barley do not carry it through consistently. Lacking a clear articulation of the relationship of such smaller groups to the larger categories of which they are instances (“fishermen,” “cab drivers,” “sweepers”), Van Maanen & Barley inevitably fall into traditional Hughesian discussion, using abstract groups like “police detectives” and “janitors” as examples of “occupational communities.”159 While the article at times seems to distinguish between “occupations” (presumably understood in Hughes’s sense) and “occupational communities,” this distinction is not discussed explicitly and is not employed consistently.

      While mostly relying on local occupational communities in their examples, Van Maanen & Barley (1984) add an important caveat: their use of the term “community” does not presume that such groups are localized. Instead, they argue:

Propinquity is… an attribute along which occupational communities vary. Certainly, propinquity may hasten and otherwise contribute to the development and maintenance of an occupational community, but it is not itself a definitional matter. Whether a particular community is geographically dispersed or clustered is an empirical question to be answered as communities are identified an analyzed (p. 298)

I see this observation as valuable and important, but not sufficient for resolving the problem of expansion across space in general. Van Maanen & Barley’s first definition of the occupational community seems to assume that the members actually know each other. They later clarify this when they contrast occupational communities to “external reference groups”:

Although external reference groups may exist for members of some occupations such as tradesmen, academics, or industrial scientists, people in many lines of work do not know people who do denotationally similar work in other settings. Police officers, teachers and fishermen know there are other police officers, teachers, and fishermen in other work settings, but they may not know them or interact with them on more than a sporadic or episodic basis. In many lines of work there are no annual meetings to attend, trade journals to read, or frequent opportunities available to meet colleagues outside the workspace who are not also members of one’s employing organization. (p. 341.)

“Police officers” thus do not constitute an “occupational community,” but rather a group defined by “denotative similarity” of work or “an extended work group” (a concept not developed further). Assuming, then that “occupational communities” are understood as small groups where the members personally know each other, Van Maanen & Barley are correct in observing that such communities may be scattered geographically. American “social theorists of… symbolic interactionist… bent” (p. 296) may be a good example of such a community, scattered by virtue of the fact that few cities have demand for more than a handful of social theorists. Unlike the policemen, they do reconvene for annual meetings, where the members have a chance to meet each other. Such communities, however, are an exception rather than the rule, and allowing for the possibility of the geographically scattered occupational communities does not help us understand how occupational communities (typically localized) relate to their “external reference groups.” The fact that American social theorists keep their community alive through annual meetings sheds little light on the relations between police work in Berkeley and New York or between software development in San Francisco Bay Area and Rio de Janeiro.

      As we consider the distinction between the local groups based on mutual acquaintance and the abstract groups based on denotative similarity of work, we may find that many attributes that Van Maanen & Barley attach to occupational communities are actually linked to those “extended” groups. For example, Van Maanen & Barley point out that member of occupational communities share specialized jargon and give an example from police talk: “We apprehended that dirtbag on a stand-up just next to my duck pond on 3rd and Main” (p. 299). While this phrase may be meaningless to true outsiders, it would be readily understandable for policemen in much of the United States, not just in “Union City,” where John Van Maanen presumably heard it. For example, we can see the term “duck pond” explained in an editorial on a website called “officer.com”: “Depending on what part of the country you’re in, the local cops will have a ‘duck pond,’ ‘cherry patch,’ or ‘cash register’ to visit when things get slow on the street” (Dees 2006). The editorial further talks about the so-called ten codes (for instance: “What am I supposed to say when I’m 10-8?” — meaning “in service”). Commenting on Virginia’s recent attempt to ban ten codes (Sheridan 2006), Dees, who served as an officer in a different state, writes: “Radio codes are part of the jargon that cops use to set themselves apart from common citizens.” Dees and Sheridan both note that ten codes vary from one place to another:

To Arlington police, “10-13” means “officer in trouble.” To Montgomery County police, the same code means “request wrecker.” Even everyday police commands can get lost in translation: In Alexandria, “10-54” refers to an alcohol sensor. For Virginia State Police, it’s livestock on the highway. (Sheridan 2006.)

At the same time, however, the use of ten codes seems to be associated with the larger group of policemen — something that “cops” do. The efforts to get rid of ten codes is of course an attempt to unify police practice, making it easier for policemen who enter the practice through different departments to work together.160

      Van Maanen & Barley argue in several places (for instance p. 295 and 341) that occupational communities must be defined “phenomenologically,” by finding out, through ethnographic work, where the members themselves draw the boundaries. This approach brings obvious benefits, but is problematic for two reasons. First, it ignores the fact that the members may draw different boundaries in different situations. While policemen may draw clear boundaries between specific departments or cliques in some cases, they are likely to identify as just “cops” in other situations. (See the quote from Dees above.) In other words, we cannot assume that member-recognized boundaries just exist — we need to look at boundary-drawing as an active process. Second, to the extent that some Van Maanen & Barley’s observations apply to the larger, abstract groups, we must remember that boundaries recognized by outsiders matter, sometimes more then those recognized by the members. A police officer would not be able to carry out many of his activities as a policeman unless he or she is recognized as such by non-policemen. The same is true (to a lesser extent) of software developers. Importantly, this is a two way relationship. The group often has a substantial power over who the public recognize as members (e.g., police officers prove their membership by showing their badges). At the same time, the outsiders’ perceptions of group boundaries most certainly affect the group’s own definitions of who is a member.

      To understand the reproduction of a practice such as software development we must look at the groups of people defined by participation in this practice, recognizing those groups as delineated simultaneously by what “social” boundaries (“objectified forms of social differences”) and “symbolic” boundaries (“conceptual distinctions made by social actors”) (Lamont & Molnár 2002, p. 168). We must also look at the symbolic boundaries as contextual and negotiated. I will now elaborate this idea using Giddens’s theory of structuration (1979).

Structuration Across Space

      To analyze the effects of abstract categories we need a theory of how they relate to practice on the ground. To develop such a theory I draw on Giddens’s theory of structuration (1979), extending it to deal with the issue of space. Following his notion of “duality of structure” and the stress on “essential recursiveness of social life” (p. 5), I see the abstract categories as descriptive of structural properties of a relatively stable (but often evolving) system of social interactions. Such categories are simultaneously constituted by what people do and are structuring their actions. The structuring effect of the categories is possible because they provide resources that actors can use to influence each other’s actions. To give the simplest example, “police officers” have the power to stop cars because the power of “police officers” to stop cars is recognized by most people. The fact that the concept of “police officers” is understood and recognized by the outsiders allows the members to generally go about their job without resorting to violence, reserving the use of force only for those isolated individuals who do not accept their authority. When violence is applied, its use is aided by the fact that those to whom it is applied cannot count on support from others.

      Software developers in Rio de Janeiro in a certain sense do the same work as software developers in San Jose. This similarity makes it possible for us and for them to talk about “software developers” as a group and about “software development” as something that they do. Once the ideas of “software developers” and “software development” become discursively available to both members and non-members, some people may seek services of “software developers” and other people may claim to be software developers. If someone’s claim to be a software developer is accepted by the potential clients, then this individual can derive income from “software development.” They can then engage in software development activities full time and gain access to better tools and more central projects. This would further strengthen their claims to membership in the category. Those who fail to convince clients that they can do “software development,” would have to either write software as just a hobby or give it up altogether. In a similar way, the individual needs to be accepted as “a software developer” by the local community of practitioners in order to draw on it for support. (To be more precise, they might need to be accepted as “a hacker” or “a coder,” since the members might use different labels from the outsiders.)

      On the other hand, the need to continuously assert membership in the abstract category creates reasons for keeping the practice synchronized with that of people whose identity as “software developers” is unquestioned — for example, the people who write software at Google. Synchronizing the practice in this case means not only synchronizing the “technique” that is applied to the relevant objects, but also displaying fluency in a system of cultural codes that would presumably make one acceptable as “a software developer” by those members. In other words, to pass as a software developer locally you need to show that you can act in such a way as to make it believable that if you were to go and talk to software developers at Google, then they would accept you as one of them. (Yet the software developers at Google are only hypothetical judges of membership.) Local representations of how the practice is carried out remotely thus become powerful structuring (and synchronizing) resources, as they give local members ability to censure lack of compliance with the “standard” practice. (As we will see later, the ability to improve upon the standard practice through innovation is in some sense the ultimate test of “central” membership in the category. However, the local innovator must be ready to defend their practice as an innovation with global significance — something that can realistically be adopted by the other practitioners and become a part of “standard” practice later — rather than a local divergence born from insufficient mastery of the standard. This sets a very high bar for innovation.)

Internal and External View of the Categories

      There are two ways to understand a category such as “software developers”, both of which can be applied recursively in Giddensian way. We can loosely call them the “internal” view and the “external” view. Both are important and I will look at the relationship between the two of them after first considering each of them individually.

      Hughes’s (1958) suggestion that an occupation can only be understood when considering its place in the societal system of division of labor represents the “outsider” view of the occupations. This view is also acknowledged (but not elaborated) in Lave & Wenger (1991), who write that communities of practice must be understood in relation to other practices. The practice of policing traffic can be understood as the activities that are undertaken to prevent certain other practices such as speeding or running red lights. In addition to being inseparable from the practices that it aims to prevent (there would be no point to traffic policing in the absence of speeding and other forms of traffic violation), the practice of policing traffic is just one of several practices that share the work of curbing traffic violations. Traffic police officers can only fine drivers for running red lights if the red lights are installed and working, yet installation and maintenance of the red lights is not something they do. The work of traffic police can thus be understood as a role in the larger system of regulating traffic. This then gives us a clue as to how we could identify “traffic policemen” in a new context: if we follow a car in Rio de Janeiro and see it stopped after running a red light, we would have good reasons to think that the person who stopped it is a traffic police officer. As this example shows, it is important to look at a range of relationships that define the practice, and not narrowly at the collaborative division of labor — the practices of running red lights or hijacking cars are as important for defining the work of traffic police as the practice of installing the lights. It also points out that we must consider the practice in relation to material objects recognizable to outsiders — such as the red lights and the vehicles.

      An internal view of the categories involves instead looking at the similarities between the members of the group that the members themselves recognize as important or even defining and of which the outsiders may be mostly unaware. American police officers speak in ten codes (see above). Software developers world wide write instructions for computers using “text editors” with names that would be unrecognizable to most outsiders (“emacs,” “vi,” “Eclipse”) and they may likely question any claims to being “a software developer” coming from someone who does not use a “proper” text editor to write code. To use Hughes’s term, the members can recognize each other by the technique that they use. They also recognize each other by what Hughes and Becker call “culture.” This culture consists of both particular ways of seeing and labeling objects, but also of particular fixed ways of talking about them — what I call “lore,” a collection of stories that circulate between the members and illustrate certain elements of collective wisdom. They may pay attention to how a particular member came to engage in the practice, considering both where they entered it (see the earlier quote from Duguid 2005), and for what reason (see chapter 2.1, “Nerds”). This identification, however, is not necessarily a matter of conscious evaluation. Brazilian software developers sometimes say that ultimately one can recognize a true developer because “you can see it in their eyes.”

      The internal and external views of the category do not always correspond cleanly to insiders’ and outsiders’ definitions. Outsiders who are looking for members of the category to perform services for them may be keen on finding people who are members of the category according to other members. (After all, anyone that they hire will end up satisfying the external definition by virtue of being hired to do the job.) Similarly, insiders may consider the members role in the external system of practices when evaluating their claims to membership.

      This dual definition of membership demands a theory of a relation between the internal and the external understanding of the category. Becker’s (1963) suggestion that culture arises from shared problems may at first glance seem to suggest the external definition as primary. Under this interpretation, we can see membership in a category like “software developers” as a matter of taking a defined role in the system of division of labor. Cultural affiliation can be then seen as simply arising from the fact that those who take this role face similar problems, both in their relations to the objects with which they work and in their relations to other practices. Such an approach, however, would see the practitioners as somehow thrown into the role all of a sudden, only then to work out some kind of shared culture. This interpretation would be contrary to Becker’s theories of learning (e.g., 1953) as well as the theory of Legitimate Peripheral Participation in Lave & Wenger (1991).

      Alternatively, we can see the categories as referring to groups of practitioners tied by shared technique and culture. Members of such “named groups” (by analogy with Hughes’s “named occupations”) collectively create for themselves roles in the social division of labor. Over time, such groups can establish for themselves new roles, which in retrospect may have little to do with the role that they played originally, and the community may eventually split. For example, electronic engineers (predominantly men) gradually took upon themselves the task of programming computers — the work that was originally assigned to human “computers” — a “subprofessional” category of people performing calculations, composed predominantly of women. Considering the fact that the first general-purpose electronic computer (ENIAC) was programmed by a group consisting entirely of women (Fritz 1996, Koss 2003), it may seem surprising that software development is today not only de facto dominated by men in almost all countries (Galpin 2003), but is also constructed internally in highly gendered terms, leaving little space for women. We must consider, however, that this change did not occur by individual men entering the communities of human “computers” one by one, gradually masculinizing the culture of human “computers.” Rather, electronics engineers, already a cohesive and highly gendered group, recognized programming work as promising and laid a successful claim to it, pushing out the female “computers.”161 This latter view, however, cannot explain how people in an altogether different location come to enter the ranks of a category such as “software developers.”

Definitions as Tools

      We can reconcile the two views if we look at the external and internal definitions of membership not as naturally constitutive of the groups, but rather as discursive tools used by members and non-members to negotiate rights to engage in certain practices. Fulfilling the role de facto is one test that individuals can try to use. The ability to demonstrate similarity to other members is another test. When those tests yield different results, the individuals may engage in negotiation as to which one is more relevant in a given context. Members of the group can also engage in “boundary work” (Gieryn 1983), trying to educate the general public as to what categories should be considered important in specific circumstances.

      It will help to look in more detail at the different types of “moves” that can be performed using such definitions as resources.162 I illustrate those moves with examples from my own research, which I expand in the later chapters. For each of the moves I consider the potential sources of resistance.

      1. Homesteading a new role. Members of an existing “named group” of practitioners may decide to engage in a new line of work, seeing this work as fit for their technique and culture, and using this fit to justify their claims to this work. They may originally engage in this work as members of their group of origin. For instance, electronic engineers may argue that as electronic engineers and makers of things like computers they are uniquely qualified to program computers. Their claims may compete with those of other groups, who may have their own arguments for why their technique is more fit for the job. If one group’s claims are accepted by those who control access to computers, the members of the group will be given a chance to work with the computers. Overtime, such people and the communities they form may develop specialized technique and an independent identity from electronic engineers.

      2. Individuals joining the group by taking a replicated role. A role may be replicated in a new place as a part of a larger replication project (in which a system of roles is replicated), creating a need for local members of the abstract group. People who take those jobs will the become members in the category by virtue of taking on the role, despite likely lacking in both technique and the culture that other members of the category share. For instance, when the Brazilian statisticians purchased an American UNIVAC in 1960 in order to replicate American census-taking practices, they created a role for people who would program it. Some individuals were selected to take this role and had to then learn the technique and culture of American UNIVAC programmers.

      Such learning presents a formidable task for the new practitioners who may be culturally isolated from their remote colleagues. Physical travel may help somewhat. (The first group of Brazilian UNIVAC programmers completed training in the United States, and American programmers were brought to Brazil to work with them side by side.) Reliance on shared background (e.g., common knowledge of standard mathematics) helps too. Finally, access to tools and objects is crucial: the presence of the UNIVAC itself (the same machine as used in the American census earlier) gave the Brazilian programmers both a claim to membership and the ability to face many of the same problems as their remote colleagues, establishing some common ground. Such resources turned out to be hardly enough, however. As I show later, many things went wrong with the UNIVAC in Brazil. While some of the problems may be attributed to the lack of mastery of the technique, it is important to consider perhaps the hardest hurdle faced by the Brazilians tasked with taming the machine: the need to take a role in a partially reproduced system of division of labor. For example, while having a “real” computer, the programmers did not have access to spare parts and had to use Brazilian-made punch cards. Perhaps more importantly, they had to mediate between the American machine and the Brazilian census organization. As a result, the census of 1960 did not get tabulated until 1975.

      3. Local individuals building ties with the group, then carving out a role locally. Around the same time, other individuals in Brazil claimed membership in a related occupation (electronic engineers) by first developing ties to remote colleagues. They did so at the time when there was no role for “electronic engineers” in the Brazilian division of labor. They traveled abroad to learn the technique and culture; some of them worked in research labs in the United States and got doctoral degrees in American universities. When they came back to Brazil, they worked to transform the local and global systems of division of labor in such a way as to create a role for “electronic engineers.” They did so by convincing other actors that a different system of division of labor was possible because Brazil had people capable of building computers, namely themselves. Doing so of course also meant overcoming many sources of resistance. The clients, such as the census bureau and the Navy, understood quite well that a handful “electronic engineers” would not be able to build computers by themselves. The engineers thus had to argue that the requisite system of economic relationships could be constructed. A number of them consequently had to put aside engineering and become policy administrators, working to create the necessary alliances between customers, investors, entrepreneurs and the state (Evans 1995). As I discuss below, such alliances worked for two decades, but fell apart eventually, largely closing the space for electronic engineers in Brazil. (See chapter 2.2.)

      It may be instructive to compare this “move” as it happened in Brazil with the process described by Saxenian (1999, 2005, 2006). The Brazilian engineers who went to study in the US are similar to Taiwanese and Indian “argonauts” in that they went abroad to the places where the practice is strong, learned the practice there and then returned to apply at the periphery what they learned at the center, having to configure local resources so at to make it possible. Compared to the “argonauts,” however, most of them spent relatively little time in the US, typically returning soon after completing their degrees. They thus likely had more limited understanding of the practice they were trying to replicate, understanding the technology without as much insight into the social workings of American industry. Perhaps more importantly, their ties to the American industry were weaker, and they largely detached themselves from it upon returning to Brazil. Their approach was also different from that of the Taiwanese and Indian “argonauts” in that they focused on replicating in Brazil a miniature version of the American practice of computer design, using knowledge they brought from the US and additional knowledge that they hoped to purchase later. The strategy put less emphasis on participation in the same system of activities as their American counterparts.

      4. Shifting to a different role when the current role disappears. Having created local space for engagement in a practice, the members can continue doing so for a while, but their ability depends on the continued existence of the role that they created, which in turn depends on other practices. If the role disappears, they will have to shift to another one that may be suitable for their technique and culture. In some cases, this new role may itself be an outcome of the same change that led to the disappearance of the earlier one.

      For example, in the early 1990s Brazilian government opened its market to imported computers, which led to quick disappearance of Brazilian computer manufacturers, closing the role for Brazilian electronic engineers. This also dramatically expanded, however, the need for software developers who could help local organizations build custom applications to be run on the now available cheap imported computers. Most electronic engineers consequently shifted to software development.

      Such a shift may involve many of the difficulties of Move 2. In Brazil, however, it was substantially easier than the ones described above. The foreign system of practices had been replicated more thoroughly by the 1990s. The clients’ suspicions as to whether the local engineers were qualified as “software developers” were also limited by the relative lack of alternatives: while the clients could buy generic computers from abroad and license or pirate some of the generic software, they often needed custom software adjusted for their specific needs, and such work had to be done locally. Bringing American software developers to Brazil was hardly a possibility, especially considering that their salaries were going up by the day. The clients had little choice but to hire local contenders.

      As Brazilian software developers soon realized, however, there was a limit to what they could do. While the role for people who would write custom software for local organizations was quickly established, developers who tried to take upon themselves the role of writing generic software for sale or many types of generic services on the web, discovered that foreign competition was getting fierce as American firms were discovering localization, while many certain key inputs were missing, the most important of them being access to venture capital. Sales of generic software and many types of web services require a large up-front investment in development and marketing, which in many cases is never repaid. They thus depend crucially on access to organizations willing to invest in highly risky ventures. Venture capitalists often also provide strategy advice to the companies they fund. (This is called “smart money” in the parlance of California startups.) Software developers in California often start their companies in Palo Alto in order to be biking distance rather than driving distance away from Sand Hill Road, where many venture capital firms have their offices. Many Brazilian software-developers-turned-entrepreneurs found themselves a bit too far from Sand Hill Road and lacking comparable resources locally. (See the story of Nas Nuvens in chapter 3.4.)

      5. Ongoing synchronization. Having established themselves as members (or tentative members) in the category, the local practitioners may have good reasons to maintain and expand ties to their remote colleagues. Such links may help them solve practical problems faced in the course of their activities, and they can often enhance their status as “software developers,” both individually and collectively vis-à-vis the abstract group. Such synchronization is particularly important if the foreign practice is seen as continuously evolving and the practitioners are perceived by their clients as being at the risk of falling behind. In this case, their membership in the practice is always tentative, and they must work to stay up-to-date with the technique and culture of remote members. In addition to synchronizing the technique and culture, they may also try to synchronize their role, working to further transform the local division of labor in such a way as to ensure that they have access to the same inputs as their remote colleagues and get credit for the same outputs. Some aspects of such ongoing synchronization appear relatively easy. Brazilian developers who have learned to read English can have access to uncountable documents that describe the foreign software technique in minute detail. Such documents also often teach the culture of the foreign practitioners, as they both express their views of the relevant objects and present specific ways of talking about the world. The local developers can also have access to some of the same computer hardware. The combination of such documents, the developers’ experience with the machines, and the fact that their clients are themselves interested in keeping up with their counterparts abroad make it possible for the developers to keep the basic aspects of their practice up to date with that of their foreign counterparts, despite the occasional faux pas such as mangled pronunciation of English terms. Moving towards more central kinds of practice, such as developing software platforms rather then custom applications presents clear challenges. (See chapters 3.2, 3.3 and 3.4.)

      As these examples show, the relationship between fulfilling a role, identifying with a group and acquiring a technique can go in different directions. Rather than logically entailing one another, roles, identity and technique each provide resources that may enable the actors to make certain moves, which in turn give them access to new resources. De facto performance of the role may provide a key to acquiring a technique and membership, while membership in the group is often crucial for entering the role. Each move requires active work and overcoming of the different sources of resistance. Those who gain access to the role may have to overcome their lack of mastery in the technique. Those who managed to pick up certain aspects of the technique may have to then overcome the challenge of finding a local role. Strengthening membership in the group may involve gradually using different resources in turn to strengthen the other.

      The first four of the five “moves” presented above are crucial for understanding why software development, in the American sense of the term, is possible in Rio de Janeiro today — and more generally how a practice can enter new places. I explore the history of those moves in chapter 2.2. The last “move” (the ongoing synchronization of practice) is crucial for understanding what Rio software developers do today — and more generally how practice proceeds after establishing a foothold in a new place. The different aspects of such “synchronization work” are explored in the other chapters.

      Steps 2, 3, 4 and 5 above all involve use of reflexive representations of a social system to structure social interactions. Such use, however, differs from the traditional Giddensian structuration, which assumes that the actors use reflexive representations of their society when structuring their interactions. Instead, the actors here use reflexive representation of a foreign social system to transform local interactions.163 While in the traditional structuration members’ mutual knowledge that things are done in a certain way normally becomes a resource for reproducing the same pattern of interaction, in his case, the members’ knowledge that things are done in a certain way elsewhere (e.g., “at Google”) creates a resource of changing how things are done.

      The importance of local practitioners’ understanding of foreign practice and social structure for local structuration makes it imperative that we look at the ways such understanding is achieved, considering among other things the specific proactive uses of communication technologies to learn more about what is happening outside. We must also be mindful of the potential gaps between members’ models and reality, and consider the ways in which the members may identify and mend such gaps.

Imagination

      It is important to consider that while local members may know a good deal about foreign practices and social structure, they do not necessarily always draw on them as a resource. The argument that a California company does things in a particular way may carry a lot of weight in some situations, but may be dismissed as irrelevant in others. This has much to do with local members’ understanding of the ways in which the foreign context differs from their own, which in turn depends on their model of the world as a whole and their place in it. For example, their view of themselves as living in “a developing country” becomes crucial in negotiations of when the foreign model applies.

      Appadurai’s (1996) notions of “imagination” and “the imaginary” (“a constructed landscape of collective aspirations,” p. 31) help us understand how reflected understanding of foreign structure is used in structuring local action. Individuals’ actions are constrained by the space of outcomes that they can imagine individually and collectively. (Individual imagination may be enough for individual action, but collective action requires collective imagination.) Reflected foreign practices and structure provide elements for construction of imaginary worlds, which may potentially become blueprints for action. Such imagined worlds may be different from reflected foreign practice: reflection concerns with how the world is (locally and abroad), imagination concerns what is possible. As Appadurai argues, collective imagination can become fuel for action. Such collective imagination is often a sine qua non of collective action.164

      I believe we must also recognize, however, the ways in which imagination may be inhibitive of action. A Brazilian proverb says that a dog once bit by a snake becomes afraid of a sausage. When looking at Brazilian software development today we must consider the ways in which one such past snake accident (the “failed” attempt to build a sustainable computer industry in 1970s and 1980s, discussed in chapter 2.2) affects today’s actions.

      Appadurai (1996) draws a distinction between “imagination” and “fantasy,” pointing out that “the idea of fantasy carries with it the inescapable connotation of thought divorced from projects and actions, and… has a private, even individualistic sound” and contrasting it with the projective and collective aspects of “imagination” (p. 7). I argue for a more subtle treatment of the boundary between the two and for recognizing the importance of “subvocal” imagination: imagined worlds that are too unlikely to be publicly presented as a plan for action or as an explanation for actions taken in the past, but which none-the-less influence action in profound ways. When discussing the belief in witchcraft in France, Favret-Saada (1980) uses the phrase “I know, but still…” to refer to the gap between what the actors “know” to be true (in the sense of being ready to defend their beliefs as rational) and what they might nonetheless act upon. Such “subvocally” imagined futures are often presented as a joke to make it easy to retract the idea without losing face. (At the same time, such imagined worlds may be discussed at length as long as the label of “fantasy” is kept on them.) This presents the ethnographer (and presumably the members themselves) with a challenge of differentiating collective fantasies used for sublimation from those that represent unspoken plans for action. (See chapters 3.2 and 3.4.)

Centers and Periphery in Worlds of Practice

      Our discussion of how a practice is replicated across space have assumed so far that we can identify places that serve as a source of practice and those that serve as its destination. Such distinction may not always be meaningful: we may find cases where practitioners in several places aim to synchronize each others practice in such a ways that we cannot easily identify the predominant direction of replication. In many cases, however, we can understand a practice geographically in terms of “center” (or “centers”) and “periphery.” In particular, the practice of software development has a clear set of centers with the most prominent one around San Francisco Bay Area (see chapter 2.2).

      Before discussing the notion of peripherality, however, we need to look more broadly at the internal structure of the “abstract” groups of practitioners. To do so, I borrow certain elements from the literature on “social worlds.” The idea of “social worlds” is associated most strongly with Anselm Strauss, and in particular his 1978 article “A Social World Perspective.” Strauss himself offers much of the credit to Shibutani (1955) and we can understand the concept better by looking at Shibutani’s discussion first. Shibutani writes that

[…] even in common parlance there is an intuitive recognition of the diversity of perspectives, and we speak meaningfully of people living in different social worlds — the academic world, the world of children, the world of fashion.

Modern mass societies, indeed are made up of a bewildering variety of social worlds. Each is an organized outlook, built up by people in their interaction with one another; hence each communication channel gives rise to a separate world. Probably the greatest sense of identification is to be found in the various communal structures — the underworld, ethnic minorities, the social elite. Such communities are frequently spatially segregated, which isolates them further from the outer world, while the “grapevine” and foreign-language presses provide international contacts. Another common type of social world consists of the associational structures — the world of medicine, of organized labor, of the theater, of café society. […] Each of those worlds is a unity of order, a universe of regulated mutual response. Each is an area in which there is some structure which permits reasonable anticipation of the behavior of others, hence, an area in which one can act with a sense of security and confidence. Each social world, then, is a culture area, the boundaries of which are set neither by territory nor by formal group membership but by the limits of effective communication. (p. 566, my emphasis)

A few things should be noted about Shibutani’s notion of “social worlds.” First, it has clear resemblance to the discussion of “culture,” “groups” and “outsiders” by Howard Becker (writing around the same time), but relies on a stronger and dangerously seductive metaphor of “worlds.” The term connotes — and Shibutani seems to largely embrace its vernacular connotations — substantial differences in perspective (or “outlook”) between those who are part of the world and those who are not. While looking at differences in outlook is often very important (and as our earlier discussion suggested), we will have to be careful to not let the metaphor of “worlds” lead us to thinking that of such difference as defining of practice, at the expense of other factors, such as the material conditions of the practice. Second, Shibutani’s concept of “social worlds” is quite broad and can be applied to virtually any group. Furthermore, any two people, no matter how similar, can be said to be living in two different “worlds” relative to some division (p. 567). If they do the same work, then perhaps one lives in the “world of women” and another “in the world of men” or, perhaps one lives in the “world” of “Generation X,” while another one is in “world” of “Generation Y.” Finally, Shibutani suggests that the reach of the social worlds is limited by effective communication. While communication is clearly important for establishing a shared outlook, we must never forget that it is not sufficient: one cannot gain the perspective of software developers (thus entering their “world”) just by talking to software developers or reading what they write. One must actually write software, and doing so requires access to more than communication.

      Strauss (1978) extends Shibutani’s concept of “social worlds” in a way that makes it much more relevant to our discussion. First, without redefining the term explicitly, Strauss focuses his discussion on a particular type of social worlds, considering “opera, baseball, surfing, stamp collecting, country music, homosexuality, politics, medicine, law, mathematics, science, Catholicism” (p. 121). Those examples are still quite diverse, but they are constrained compared to Shibutani’s examples (which include “the world of children” and “the social elite”). Strauss’s main focus is quite clearly on the worlds that are similar to the abstract groups of practitioners that I have discussed earlier. Strauss then introduces three ideas that further constrain the meaning of “social worlds” to groups of practitioners: he suggests that each world has specific activities (one or more of which can be considered primary), that the activities occur in specific sites, and are aided by specific technologies and organizations. Note that it would be hard to think of “the world of children” (one of Shibutani’s examples) in terms of its “primary activity” or “technology.” If social worlds are understood as defined by activities and technologies, we can map them to some extent onto abstract groups of practitioners discussed above, with an important caveat that Strauss’s social worlds are broader, as they include not only those who actually engage in the world’s activities, but also those who are “associated” with those activities more loosely, and the activities themselves are understood more broadly. Strauss himself does not offer clear examples of such extensive analysis of social worlds, but Becker’s Art Worlds (1982), which Strauss cites in Strauss (1982), can be read as providing such an example. In this book Becker attempts to describe the different roles involved in production of different kinds of art. In case of film-making, Becker looks at the producers, camera operators, focus-pullers, janitors, film critics, and many other roles. Strauss’s discussion (1978) seems to allow for such broad analysis.

      Strauss (1978) also extends Shibutani’s analysis with a more sophisticated analysis of boundaries, introducing a notion of “authenticity” and “authentication” (p. 123), pointing out that some members of the world are seen as “more authentically of that world, more representative of it” (p. 123) and raising the question of “who has the ‘power’ to authenticate? and how? and why?” (p. 123). Combined with the notions of legitimate “sites,” those questions become essential for our analysis.

      While Strauss himself does not analyze the role of space in much detail, other authors working in the same traditions do. Levine (1972) uses the term “Meccas” to refer to places that carry tremendous power in a social world, both in terms of practical ability to coordinate resources and in terms of their symbolic power, as sources of legitimation and arbiters of membership. Looking at the Chicago art world Levine points to New York as such a “Mecca,” arguing that succeeding in New York is crucial for communicating status in Chicago. As one of Levine’s interviewees says: “If you want to show in Chicago, you must move to New York” (p. 298). Local processes thus orient themselves in relation to the Meccas of the social world. As I show later, the software world similarly has easily identifiable “Meccas” on the West Coast of the United States (plus perhaps a few more places such as Boston, Helsinki, and Bangalore). References to those areas come up frequently in conversations with Brazilian software developers. (Levine says that New York’s predominance has much to do with economics: the city is home to a large number of wealthy buyers. The large number of buyers allows for many galleries, and as a result also a large number of knowledgeable and discerning buyers. The best Chicago artists consequently sell their work to New York and often move there. This means that ability to sell one’s work in New York becomes crucial to gaining status in Chicago.)

      Strauss’s concept of “social worlds,” however, suffers from a number of problems. First, the concept of “social worlds” still lacks clear boundaries, making it hard to know what is and what is not a social world, and potentially letting us see one in any group of people. Second, Strauss moves back and forth between his analog of what I called “internal” and “external” definitions when discussing abstract groups of practitioners. Shibutani’s “worlds” are unambiguously a matter of individual perspective — a view that Strauss does not explicitly disclaim. Strauss’s own discussion, however, focuses (1982) sometimes on sources of legitimacy perceived by other members and “a collective definition that certain activities are worth doing, and ‘we’ are doing them” (p. 174), and sometimes on the perceptions of outsiders. The relationship between those is left unexamined. Additionally, like Abbott (1988), Strauss often assigns agency to “worlds” and “subworlds” without explaining how such agency relates to the individual agency of its members.

      Considering those difficulties with Strauss’s analysis, I do not incorporate the notion of “social worlds” directly, but rather develop a similar notion of “worlds of practice” defined within my theoretical framework as presented in this chapter and in chapter 1.1, which aims to be narrower than Strauss’s social worlds, but sufficiently similar so as to allow me to graft some of the most important insights that Strauss adds to our discussion, in particular the importance of “sites” and “authentication.”

      We can say that a practice such as “software development” defines “a world of practice” (understood as a type of Strauss’s “social world”), to the extent that people who engage in it share certain collective definition of that practice. Understood this way, “a world of practice” is a much narrower concept that “a social world.” First, “worlds of practice” are understood as wide in their reach, operating at either global or national scale. Second, they refer to collectivities built around practice as defined earlier. I do not, for example, extend this notion to Shibutani’s “world of children.” Third, I do not aim to include everyone associated with in any way with a particular set of activities (which is the intention of Becker 1982). Just like every world has a set of “primary activities” (or we could say “primary practices”), every practice has its primary social worlds. In other words, while software developers participate in many Straussian social worlds, the practice of software development plays an auxiliary role in most of them. In my later discussion of the “software world” I focus specifically on those people who develop software, with less attention to those who use the software, market it, or make computers on which software is built.

      Thinking of a practice as forming “a world of practice” allows us to ask important questions about the geography of that practice, considering what sites are central to that practice and how such sites are implicated in authentication of individuals as members. We must be careful, however, to not assume that members who work in central sites necessarily act as arbiters of authenticity directly. Instead, authenticity is often judged separately in each site, and central sites are important mostly because they serve as models of the practice.

      In the world of software, for example, there is currently broad consensus that the most important work is done today in two small areas on the West Coast of the United States: the Silicon Valley in California (the eighty mile long area between San Francisco and San Jose) and Seattle/Redmond area in the State of Washington State. (This co-exists with the ideology of irrelevance of geography.) At such “centers,” the local community’s authenticity is rarely questioned and it suffices for an individual to focus on finding his place in the local community, without needing to worry where this community fits in the larger world. We can contrast such centers with peripheral sites. At the periphery, the global status of the local community is questioned regularly. The individual involved in a social world at its geographic periphery thus has to decide whether to cast his lot with the local community or to seek ties with those parts of the social world that lie outside. (This is of course not a perfect dichotomy — as we will see in the case of Brazilian software developers, all participants have to maintain their ties to both the local community and the outside world. They only choose what ties to focus on.) To understand what is happening, we therefore must consider side by side at least three entities: the individual, the local community (with all of its factions) and the larger world with its central sites. By understanding what drives the individuals to seek direct links with external parts of the world we will come to understand how the local community synchronizes its practice with the rest of the world.

      Considering the peripheral members also brings into focus the interface between the world of a particular practice and the broader local society. Strong worlds can overtime transform the society around their centers, making it easier for the members to move back and forth between their world of practice and the mainstream society. For example, while the “nerd” identity associated with software developers was seen in United States as somewhat unmanly in the past, it has been partly incorporated into “hegemonic masculinity” around 1980s and this process continues today (Kendall 1999). Due to this historical work, a software developer working in the Silicon Valley today rarely experiences conflict between his identities as “a man” and “a developer.” This integration is perhaps most visible in the late 1990s movies like “The Matrix.” Software developers working in Rio de Janeiro, on the other hand, operate in a place where different forms of masculinity are the norm. Reconciling the identities of being a man and a nerd is thus substantially harder in Rio de Janeiro than it is in San Francisco. The conflict between the two identities thus might be easier to observe there.

      A world of practice may also be at peace with the mainstream culture at the center because aspects of that mainstream culture are often incorporated into the culture of this world as basic assumptions. Peripheral members, on the other hand, again have to face the contradictions between the demands of the social worlds and those of their local mainstream society. This effect is perhaps easiest to see with language: software developers in California can perform all of their daily activities in one language, while their Brazilian counterparts must switch between the language of the software world (English) and the language of the local society (Portuguese). The choice of language may become important as a marker of allegiance to one or another group, as we will saw in the previous chapter and will see again in chapter 2.1.

      This observation has an important consequence for innovation that occurs at the center vs. at the periphery. Practices and knowledge generated at the center are mobile from birth. While such practices and knowledge may be inextricably tied to local culture and context, this fact does not provide insurmountable problems, since the rest of the world is typically ready to accept such practices on those terms. A book on software development written in California in English does not need to be translated to become successful world-wide: the author can count on the potential readers to either learn English or struggle through the book with a dictionary. Practices and knowledge generated at the periphery, on the other hand, have little change of success outside their local context unless they are actively disconnected from it. In other words, central actors can “disembed” their knowledge using the simplest strategy available, leaving others the hard work of re-embedding it at the periphery. Peripheral actors must perform the most thorough disembedding, to make re-embedding at the center a trivial task. In doing so, they might have to forgo the needs of local users, as we will see in the case of the Lua programming language (chapter 3.2, 3.3).

Reinforcing and Challenging the Centers from the Periphery

      The actions of peripheral members might affect the global structure of the social world in other ways as well, simultaneously re-enforcing and undermining it. Latour (1988a) points out the intellectual gains that arise from being at the center of a global network where knowledge is brought from around the world:

The zoologists, in their Natural History Museums, without travelling more than a few hundred meters and opening more than a few dozen drawers, travel through all continents, climates and periods. They do not have to risk their lives in those new Noah’s Arks, they only suffer from the dust and stains made by the plaster of Paris. How could one be surprised if they start to dominate the ethnozoology of all other people? It is the contrary that would indeed be surprising. Many common features that could not be visible between dangerous animals far away in space and time can easily appear between one case and the next! The zoologists see new things, since this is the first time that so many creatures are drawn together in front of someone’s eyes; that’s all there is in this mysterious beginning of a science. (p. 225.)

Latour points out here that the foundation of European science lies in the massive increase in basic knowledge of the world made possible by its central position in a colonial empire. (Latour refers to such places as “centers of calculation,” pp. 215-257.) Such accumulation was not limited to the animals, but to cultural artifacts and the indigenous individuals themselves. While a few centuries ago colonial subjects were often brought to the center by force, today many go there of their own will, and it is usually the most talented of the peripheral members that gather at the center.

      At the same time, peripheral recreation of the local practices can in the long term create alternative centers. While most parts of the world are unlikely to ever occupy this role, some of them can rise to rival the earlier centers.165 Bangalore, for example, has risen to be a minor Mecca of the software world, though it is far from becoming a rival of Silicon Valley, so far playing a clearly subordinate role. (The best work available to software developers in Bangalore today is provided by companies based in the US.) Brazilian developers and economists might wonder why India and not Brazil has risen to play second fiddle, but it is still California that rules their imagination.

      Additionally, such success might require that the local members of the world of practice create an enclave that is separated from the local context. The success of Bangalore as a software hub, for example, might be attributed to the city’s lack of commitment to any local language and willingness to adopt English as the working language.166

      Such places may also be handicapped by lack of proximity to the centers of other worlds. San Francisco is a Mecca of a number of worlds, which reinforces the position of each of them. On the other hand, while Helsinki has become somewhat of a Mecca of the mobile world due to Nokia, this hardly means that mobile developers in Brazil would consider learning Finnish. This means that while considering the geography of individual worlds of practice, we must keep in mind the politico-economic structure of the world as a whole. Rio’s position in the software world in many ways corresponds to Brazil’s “semi-peripheral” position in the world economy as a whole, as suggested by Cardoso (1972) and Evans (1979).

Parts 2 and 3

      I presented in this chapter an abstract discussion of the process through which a technical practice that originated in one place is reproduced in others, and the asymmetric geography of practice that results from this process. It is in this context of a partially reproduced system of relationships and peripheral position in the larger world defined by the practice that we must look at the work done by software developers in Rio de Janeiro. In Parts 2 and 3 I try to put more flesh on this model. Part 2 explores in more detail the history of the individual and collective entry into the world of software. Part 3 looks at the different ways in which peripheral actors may organize the local and global resources available to them.

Notes