Google: Meta-communities of practice in the code-sharing commons

The extraordinary proliferation of software projects in the last two decade defies easy analysis. Software saturates everyday life (Dodge 2011). This research project uses data generated as people work on online code repositories and use programming Q&A sites to analyse the diversity and homogeneities of programming practice. While much recent debate around software, and the digital economy more generally, has attended to issues of intellectual property associated with open source or free software, this project aims to construct a new evidence-base for debate on these and other issues. It treats programming as a set of practices amenable to social scientific analysis both in terms of common patterns (patterns of coordination, development of group structures) and in terms of the migration and recombination of these practices across different settings. The aim of the research is to bring to light the diversity and sameness of practices within what we call the meta-community of programming practices. The term 'meta-community' refers to the dispersed yet linked communities of practice that populate contemporary software cultures.
Our research will draw on two major kinds of publicly accessible data produced when people make or code software together. The first kind of data comes from code repositories. Online code repositories support the collaborative production of software. A second main kind of site addressed in this project complements the first: the programmer question and answer site StackOverflow. This site collaboratively filters a huge range of very specific questions about how to write code, and how to use existing software libraries and other pieces of code. It is widely regarded as the primary site of online help for programming and software-development problems.
The major repositories such as Github, SourceForge, Launchpad, bitbucket and Google Code host millions of projects. These projects display very diverse shapes and forms, and an extremely wide range of topics, interests, investments, commitments and uses. Online code repositories supply extensive and fine-grained data on the projects, community membership and a range of actions associated with coding. They also publish the source code for the software itself. While this data is published largely in order to facilitate project management and coordination, we see the repositories as a source of rich data on how software is made in practice, and as exemplary sites to investigate different practices associated with the digital economy.
Our research will involve writing scripts to gather data from these sites, and then organising this data in a large relational database. We will focus on data relating to project topics, data stemming from certain kinds of actions (such as 'commit-actions'), data describing overlaps or dependencies between piece of code ('shared dependencies), and data derived from queries about how to write code ('practice intensity'). In order to integrate and find patterns in these datasets, we will draw on well-documented techniques of machine learning. Although these have been relatively little-used in the social sciences, we see them as an important avenue for exploring heterogeneous datasets.
We plan to disseminate the research in novel form. We will bring together a snapshot of the database, the scripts used to generate, analyse and visualize the data, and our written analysis in a virtual machine that can be downloaded and run on most computers. Virtual machines (platform-independent software environments) have recently appeared in the genomic sciences as a way of addressing problems of reproducibility in data intensive research (Encode Consortium 2012) By distributing data, code, visualization and analysis in a coherent, independent package, perhaps for the first time in contemporary social science practice, we will facilitate reproduction of the analysis and exploration of data by interested users inside and outside academia.

The project focuses on ways of tracking practices through datasets generating during programming practices such as writing code. Given the timespan and close focus on methodological development, immediate beneficiaries of this research are located within academic or near-academic settings in the social sciences.
We have identified several non-academic beneficiaries. They include practitioners of software development in various settings ranging across industry, government, commerce, media, finance, and innovation settings. The other main beneficiaries include industry advisory groups such as the BCS Open Source Software Specialist Group, and the UK-based Open Source Consortium. Software and IT- industry analysts might also benefit from our research. Finally, computer science and software engineering disciplines are academic beneficiaries through which the research can have indirect economic and social impacts.
We regard software developers of all varieties as important beneficiaries. The research will have much significance for programming communities. Analyses of the popularity of programming languages attract much attention in online programmer forums and newsites (extensive coverage of TIOBE programming language index (TIOBE Software 2012) on Slashdot.org, a newsite widely read by programmers and software developers). This research will offer a more richer description of programming practices over time. We plan to present intermediate results of the analysis at industry software conferences such as the annual PyCon conference. As various results, visualizations and summaries of this project are published on the project research blog, the study should attract sustain social media attention and interaction with us via blog comments, and Twitter conversations. Our practices of dissemination (virtual machine via FTP and USB sticks, presentations at user groups and industry conferences, online documentation in blog, use of code repository for our code) are, moreover, designed to engage and to sustain interactions with these beneficiaries.
A range of industry and policy advisory groups are public beneficiaries. Groups who promote open source software development, or provide advice on software-related activities to government and industry, are potential users of this research. In the software industry, as well as in various media and scientific settings, software development is a core innovation process. In its analysis of the diversity of programming practices, as reflected in code repositories and programming Q&A sites, the research will offer a uniquely comprehensive and comparative analysis of the importance and relevance of specific software practices (e.g. programming languages, computing platforms, application domains, etc.). It will also furnish an evidence-base for the development of policy advice in many areas affected or dependent on software development. While we do not see ourselves as delivering the kind of analysis supplied by industry analysts (e.g. Gartner Consultants or RedMonk), there are potential overlaps between their interests and the knowledge produced here.
The research is in interdisciplinary dialogue with computer science and software engineering. These disciplines can have a more immediate impact on software development practices, partly through their training-educational activities, and partly through the pursuit of their own research priorities. In relation to users in academic disciplines interested in the domain of software development (software engineering), we think that our research could potentially open up different training priorities, and different perspective on what they teach. The research might also identify interesting research needs and problems that could be addressed by the development of software tools, or other technical innovation. These kinds of impacts are most likely to occur outside the time-frame of the funded research.