Article Preview
TopIntroduction
Academic interest in the OSS movement has centered upon identifying what makes OSS special in terms of governance, organization, and the OSS process itself (Krogh & Hippel, 2006). As OSS has moved “from an academic curiosity to a mainstream focus for research” (Crowston, et al.. 2012) driven by growing user awareness from consistent uptrends in the download volume, this research inquires - Do the number of developers contributing to Free and Open Source Software (hereafter OSS) projects assist or hinder project download volume? Although it is stated that many “developer” eyes looking at the code makes bugs shallow (Raymond, 2000), there is little evidence about how the number of developers contribute overall to OSS projects.
In an OSS environment dominated by various beliefs, norms, and values (Stewart & Gosain 2006), the lack of systematic governance and participant developers’ self-selection to various OSS projects offer an interesting dilemma. With no institutional control assigned via governance, the OSS environment remains an amorphous mix of various developers with a wide spectrum of motivation and involvement. Unlike organizations that can control every resource allocation, OSS projects rely on resource self-allocation. Will such self-regulated resource-allocation lead to continuous improvement, diminishing marginal returns or something else? Specifically, is there a limit to the number of developers after which adding more and more developers loses its effectiveness and perhaps even becomes detrimental to a OSS project?
Raymond’s (2000) viewpoint that OSS, as a bazaar, greatly accelerates quality control by a large, voluntary OSS developer groups with “many eyeballs” through increased debugging and code evaluation and contrasts favorably with the cathedral approach in closed-source software. The idea is that an open-sourcing environment based on volunteers allows many developers to review OSS project’s code – thus increasing the probability of finding a bug. We contend, as an alternative, that over-allocated eyeballs (as developers) may limit the best allocation and use of resources where the “availability of source makes developers and users discount the necessity of testing that is common in commercial software” (Seltzer, 2014).
Expectations vary about outcomes between closed-source and the OSS. For example, source code and web forum support posts are generally considered non-rivalrous which creates an expectation of indefinitely increasing marginal returns. Day and Dowrick (2013) find an asymptotic function exists in instances of rivalrous and non-rivalrous resources combinatorial allocations. Therefore, even if some rivalrous resources exist within OSS projects, it would be expected that marginal contribution of developers would remain positive. This paper contends and empirically supports the contention that regardless of the non-rivalrous nature of OSS resources, OSS project download volume may diminish in the presence of over-allocated developers.
In software-development projects, Brook’s (1995) Law remarks that adding more manpower to a project is unlikely to have positive results and is actually likely to be detrimental. We investigate whether a similar corollary may manifest itself in software-development projects that may in fact suffer from a diminishing marginal value of developers that negatively impacts project download volume over time. This study intends to surface an inflection point in tracing the relationship between the (i) number of developers (ii) QC (Quality Control) and (iii) OSS download volume – confirming the diminishing returns to quality and project download volume as the number of developers rise beyond a certain magnitude in OSS projects.
In developing and proposing a model for measuring the quality of the output of OSS projects anteceded by the number of developers, the paper is organized as follows: We begin with an overview of the concepts and definitions pertinent to our model development. We then test this model by analyzing more than 12,000 OSS projects from the popular SourceForge.Net incubator. We discern that, in fact, a threshold does exist whereby the marginal contribution significantly diminishes project quality negatively affecting a project’s download volume. Lastly, we discuss the implications of this threshold in terms of impact and significance to OSS projects.