Big Difff, Granularity, Incoherence, and Production in the Github Software Repository

Abstract

This chapter will discuss the way in which Github, one of the largest dynamic repositories of software online, can be seen to operate as a mode of archive which in turn re-engineers the question of what an archive is. In very simple terms, Github is a place where software is stored online and from which it can often be downloaded. More expansively, it provides a sense of the archive as simultaneously a site of fine-grained analysis and of incoherence, of storage and of production. To get to Github, we need to start with Git, a ‘source code management’ (SCM) system designed by Linus Torvalds in 2005. 1 Git was initially based on the characteristics of a file storage system familiar to its author as the initiator of the Linux aspect of the GNU/Linux operating system. 2 Whilst it claims to be ‘a stupid content tracker’, 3 in practice Git is a highly sophisticated, decentralized, and distributed way of writing code in groups on scales ranging from an individual to that of large organizations. Git encourages branching or multiple versions of the same project at the same time and provides many different ways of merging, tracking, duplicating, and integrating code repositories distributed across many developers. It facilitates and encourages copies and variations as well as the tracking and auditing of changes in almost any kind of digital data. Since 2007, Github.com – a separate organization – has served as a largely public host platform for Git repositories or ‘repos’. It has encouraged software developers and programmers to store, work on, and retrieve the source code and texts associated with software projects on many scales, again ranging from individuals to large organizations. It has augmented the many operations afforded by Git with ‘social coding’ affordances such as ‘starring’, ‘watching’, the distinctive ‘pull-request’ mechanism, various more formal organizational arrangements (teams, organizations, etc.), and visual descriptive devices (graphs in particular). Github has grown rapidly since 2007 to become perhaps the most important online code repository of the moment, hosting around 10 million projects in total with several million people contributing to them, albeit with widely varying levels of activity. We might understand Github as the formal enterprise that organizes – and somewhat ironically, centralizes – the informal de-centralized organization of Git. Github itself publishes much data about the growth of repos. The public legibility of platform dynamics is typical of contemporary software-mediated culture: things are made to be readable by many. Github.com also produces and encourages the production of various forms of visualization and tabulation of what goes on there. To illustrate this legibility, we could choose important or famous repositories on Github – the Linux kernel, for instance, still led by Linus Torvalds, is a much-vaunted FLOSS project that has become economically and technically central to the development of the Internet – and analyze the flows of meaning, texts, and readers/writers connected to that repository. 4 Relatively quickly, individual contributions could be analyzed, and we could begin to characterize the composition of the group of people who keep this important software object working and up to date. But this work is largely already done by Github.com itself. 5 Indeed, the site is characterized by a high degree of granularity of the data it holds. This is understood to mean the availability of multiple kinds of highly detailed, and to some degree tractable, information of the processes, material, and actors it gathers. Since Github is notable for the ‘socialization’ of software production, in which the above-mentioned social media forms are built into the archive, there is, in turn, a deep integration of quantification into the working processes of the archive.