July 02, 2026
Open source software (OSS) is not homogeneous. A project’s purpose, governance, and funding shape how its community forms, who contributes, and how the software is maintained, yet empirical research often samples OSS broadly and reports findings as if they held for open source as a whole. We argue that OSS comprises distinguishable sub-genres, and that the sub-genre a study samples bounds how far its findings generalize. Using a light, multi-source review that screens 3,925 unique papers, we synthesize a typology of fourteen OSS sub-genres, from well-studied ones such as community-driven, company-backed, foundation-governed, research and scientific, and open source for social good (OSS4SG), to under-studied ones such as multi-company co-opetition, protestware, and open-source appropriate technology. We place the sub-genres in a framework that records each one’s primary driver, governance, and funding, with its maturity in the literature and representative projects, and we present a research agenda whose central question is whether findings established on one sub-genre transfer to others. The contribution is the typology and the agenda rather than a complete census, and we mark the sub-genres whose empirical support is thin.
Open source software (OSS) is not one thing. Empirical research nonetheless samples it as though it were. Studies draw a set of popular GitHub projects, fit models of contributor retention, onboarding, or code quality, and report the results for open source in general. The implicit assumption is that a community Linux distribution, a single-vendor database, a university research library, a humanitarian health-records system, and a one-person utility differ only in size and popularity. They do not. They differ in why the project exists, who decides its direction, how the work is paid for, and how the community behaves.
Recent evidence makes the difference concrete. A comparison of mission-driven open source for social good (OSS4SG) and conventional OSS reports different community structure, contributor retention, and code-quality management, with OSS4SG communities retaining contributors (“sticky”) and conventional projects attracting many who do not stay (“magnetic”) [1]. The newcomer-to-core pathway also differs by sub-genre, and contributors in OSS4SG reach core status at higher rates and through more than one route [2]. Even the way a newcomer first joins, whether through a mentorship program, a hackathon, or independently, is associated with how long they stay [3]. When two sub-genres diverge this far, a retention model or an onboarding program validated on one of them need not hold for the others. Coarse metrics such as stars and commit counts hide the structure that decides whether a finding generalizes.
The distinction matters in practice as well. A developer deciding where to contribute gains from knowing what kind of project they are approaching, because its governance, funding, and community behavior set what a first contribution takes and how a newcomer is received. Social barriers at the first contribution already push newcomers to abandon projects [4], and the path into a single-vendor product differs from the path into a volunteer distribution or a research library. Sub-genre bears on sustainability too. How a project is funded and who maintains it, and so how resilient it is, ranges from firm-employed teams to the few unpaid volunteers behind widely used infrastructure [5], [6]. Naming a project’s sub-genre makes these differences explicit instead of leaving each contributor and each study to rediscover them.
This paper makes three contributions. First, we present a typology of fourteen OSS sub-genres (Section 4), organized by who drives a project and to what end. Second, we place the sub-genres in a framework that records their driver, governance, and funding (Section 5). Third, we present a research agenda that treats sub-genre as a variable to report and control for in empirical OSS studies (Section 6). The paper is short by design. We mark the sub-genres whose evidence is thin as targets for future work.
Prior work classifies open source along single axes. [7] contrasts a volunteer “OSS 1.0” with a commercially co-opted “OSS 2.0”. [8] separates community open source from single-vendor commercial open source, and [9] contrasts community-managed projects with firm-involved hybrids. Other schemes classify by the firm-community relationship [10], by governance configuration [11], [12], by business model [13], by contributor motivation [14], or by structural growth dynamics, as in the stadiums, clubs, federations, and toys of [15]. A broad survey of free and open-source development covers the literature across topics from individual contributions to governance and community dynamics [16].
These schemes are valuable but partial. Each cuts the space on one dimension, and most resolve into a binary or a few types that center on commercial form and leave little room for non-commercial and public-interest projects. The typology presented here differs in two ways. It organizes OSS by who drives a project and to what end, an axis that the commercial-versus-community binaries approximate but do not fully separate, and it spans fourteen sub-genres that include forms the earlier schemes collapse or omit, such as multi-company co-opetition, government and civic OSS, open-source appropriate technology, protestware, and critical digital infrastructure.
We conducted a light, multi-source review rather than an exhaustive systematic one, and we release the search-and-screen script so the corpus can be regenerated. We queried two scholarly indexes, OpenAlex and arXiv, with 46 queries grouped into fifteen facets, one facet per candidate sub-genre plus a discovery facet aimed at sub-genres we had not anticipated, such as “taxonomy of open source projects”. The search returned 5,771 records. We normalized them to a common schema and removed duplicates by DOI or title, which left 3,925 unique papers. We then screened the papers with a rule that keeps a paper when its title or abstract both refers to open source and signals that the work defines, classifies, or studies a kind of OSS project or community. Of these, 399 papers passed. Coverage across the fourteen sub-genres ranged from 8 papers, for critical digital infrastructure, to 76, for OSS4SG, and reflects query reach rather than the size of each field. We grouped the screened papers into the sub-genres below. The search-and-screen design follows the bibliometric and systematic-review practice used in prior work [17], [18].
We treat the resulting typology as a working hypothesis to be confirmed, split, merged, and extended, not as a settled ontology, and Section 7 states its limits.
| Sub-genre | Primary driver | Governance | Funding | Maturity | Examples |
|---|---|---|---|---|---|
| Community-driven | Intrinsic: learning, reputation, ideology | Meritocratic, peer-elected | Volunteer, donations | Established | Debian, Arch Linux, GNU Emacs |
| Company-backed OSS | Commercial: revenue, market control | One firm controls | Firm-employed staff, product revenue | Established | Elastic, MongoDB, GitLab |
| Foundation-governed | Neutral multi-party collaboration | Foundation rules, vendor-neutral | Member dues, corporate sponsorship | Established | Apache HTTP, Kubernetes, Eclipse |
| Multi-company co-opetition | Strategic shared work among rivals | Shared across firms | Competing firms’ staff | Emerging | OpenStack, Linux kernel |
| InnerSource | Internal reuse, efficiency | Trusted committers within a firm | Corporate, internal | Growing | HP, Philips, Ericsson |
| OSS for Social Good (OSS4SG) | Societal benefit | NGO, nonprofit, community | Grants, donations, NGOs | Growing | OpenMRS, DHIS2, Ushahidi |
| Government and civic | Public mandate, transparency | Publicly accountable | Public funds, procurement | Growing | Decidim, X-Road, GovStack |
| Educational | Learning and social good | Faculty or instructor led | Universities, course staff | Growing | Sahana Eden, LibreFoodPantry |
| Open-source appropriate technology | Basic needs, sustainability | Practitioner and academic networks | Grants, NGOs | Emerging | RepRap, WikiHouse |
| Hobbyist and solo | Personal enjoyment, learning | None, single maintainer | Unpaid, occasional donations | Established | left-pad, micrograd |
| Protestware | Political or economic protest | Maintainer acts unilaterally | Unpaid | Emerging | node-ipc, colors.js |
| Research and scientific (RSE) | Research support, reproducibility | Laboratory or PI led | Research grants | Established | Astropy, scikit-learn, LAMMPS |
| Open hardware and open data | Reproducibility, maker, civic data | Domain consortia, communities | Grants, community | Growing | Arduino, OpenStreetMap |
| Critical digital infrastructure | Public good, technical interest | Informal, few maintainers | Under-funded, ad-hoc sponsorship | Growing | OpenSSL, curl, log4j |
We organize the sub-genres by their primary driver, that is, who starts and sustains the project and to what end. The axis is not perfectly clean, since many projects span more than one sub-genre, but it captures the main source of variation. Table 1 summarizes all fourteen sub-genres by primary driver, governance, and funding, with maturity and examples. The subsections below define each sub-genre and cite representative work, and we mark the ones whose dedicated literature is still thin as emerging.
Company-backed OSS is owned and primarily developed by one firm that earns revenue from the software, through an open-core model (a free core paired with paid proprietary extensions), dual licensing, or a hosted service, while the firm controls the project’s direction. [8] defines the single-vendor commercial form, and later studies analyze how firms participate in and govern such projects [19], [20]. Multi-company co-opetition is an emerging sub-genre in which firms that compete in the market jointly develop shared infrastructure. [21] examined cooperation among competing firms in OpenStack [22], and recent work characterizes how such collaborations are governed [23]. InnerSource applies open-source practices inside a single organization, adding open contribution and trusted-committer review to internal development for cross-team reuse [24]–[26].
Community-driven OSS is the classic peer-production model, built by volunteers who contribute from intrinsic motivation and where authority follows merit rather than ownership. [27] established the economic theory of commons-based peer production, and empirical studies describe the social and authority structure of projects such as Debian [28], [29]. Foundation-governed OSS places a project under a vendor-neutral nonprofit, such as the Apache Software Foundation, the Linux Foundation, or the Eclipse Foundation, that holds the trademark and intellectual property and enforces neutral rules so that competing firms can collaborate [30]–[32].
Open source for social good (OSS4SG) comprises projects whose primary purpose is societal benefit rather than commercial value, such as health-records systems and crisis-response tools, and many appear in digital public goods registries. [33] studies the motivations and challenges of contributing to OSS4SG, and later work examines its contributors and community dynamics [1], [2], [34]. Government and civic OSS, including software designated as a digital public good, is commissioned or mandated by public institutions under transparency and procurement constraints [35]–[37]. Educational OSS is produced inside teaching contexts, with students contributing for course credit. The Humanitarian Free and Open Source Software (HFOSS) tradition is its archetype, pairing software-engineering education with socially beneficial projects [38]–[40]. Open-source appropriate technology (OSAT) is an emerging sub-genre that extends open-source practice from software to hardware and design knowledge meant to meet basic needs in resource-limited settings. [41] articulates OSAT and demonstrates it in practice [42], [43].
Hobbyist and solo OSS consists of projects that one person builds for enjoyment or to meet a personal need, with no formal governance and a low truck factor, meaning that few contributors would have to leave for the project to stall [6], [44], [45]. Protestware is an emerging sub-genre in which a maintainer alters or sabotages a package to deliver a political or economic message, so that the defining act is conflict rather than feature work [46]–[48].
Research and scientific software (RSE) is built inside research to support and reproduce studies, usually by domain scientists rather than trained software engineers, and it is funded by grants. It is at once a research output and a research instrument [49]–[51]. Open-source hardware and open data apply open-source practice to non-software artifacts, including physical designs and public datasets [52]–[54]. Some projects release an open dataset as the primary artifact, such as in computer-vision-based fitness analysis [55]. Critical digital infrastructure OSS comprises components that much of the software ecosystem depends on indirectly, such as OpenSSL and curl, yet that a few volunteers often maintain. [5] named this sustainability problem, and later work formalizes it as underproduction, where the labor a component receives falls short of how widely it is relied upon [56]. Studies of the npm ecosystem also show how a few maintainers’ accounts reach much of it [57].
Table 1 shows two patterns. First, the primary-driver axis does not fix the other dimensions. Foundation-governed OSS and multi-company co-opetition both involve competing firms, yet they differ in where authority sits. Second, the maturity column shows where evidence is thin. Several sub-genres that matter in practice, namely multi-company co-opetition, protestware, and open-source appropriate technology, rest on a few studies each, against the deep literatures on community-driven, company-backed, and foundation-governed OSS and on research and scientific software.
This is the question the typology raises. A retention model built on company-backed OSS, an onboarding program validated on a large community project, or a code-quality heuristic drawn from popular repositories need not transfer to research software, humanitarian projects, or solo-maintained infrastructure. Prior work already shows that contributor retention, the newcomer-to-core pathway, and the effect of entry events differ between OSS4SG and conventional OSS [1]–[3]. How far this difference extends across the other sub-genres is open. Research and scientific software is a concrete next case, since its contributors work to research calendars and academic careers rather than product cycles, and whether community-structure findings from conventional OSS hold there is untested.
Critical digital infrastructure is the clearest case, a sub-genre the whole ecosystem depends on yet one defined by few maintainers and limited funding [5], [56]. Multi-company co-opetition and protestware are similarly consequential and similarly thin. A sub-genre-aware research program would direct effort toward these cases.
A study should report the sub-genre it samples, and control for it where possible, as studies already report programming language or project size. Tooling can help. Methods that recover intent from artifacts, such as generating user stories from source code with large language models [58], point toward classifying projects into sub-genres at scale, which would make sub-genre-stratified analysis routine. Classification at scale would also serve practice, letting a contributor see what kind of community they are joining before investing effort, and letting maintainers and funders find the sub-genres where support is thin.
The review is light rather than systematic. It covers English-language and well-indexed venues, so we may have missed sub-genres visible mainly in other languages or in grey literature. The sub-genres are not mutually exclusive. Kubernetes is at once foundation-governed and multi-company co-opetitive, and OpenMRS is at once OSS4SG and educational, so the sub-genres overlap rather than partition the space. The maturity labels reflect the literature we retrieved rather than a citation count, and they can understate a sub-genre whose evidence is recent or scattered. The per-sub-genre screened counts reflect the reach of the queries, not the true size of each field. The dimensions are deliberately coarse, and we leave finer stratification, such as by application domain, to future work.
Open source is plural. Treating it as one thing lets coarse metrics hide differences in purpose, governance, funding, and community behavior, differences large enough that a finding from one sub-genre need not hold for another. We have presented a typology of fourteen OSS sub-genres, placed them in a comparison framework, and proposed a research agenda whose first task is to test how far existing findings generalize. The typology is provisional. Its value is in making the variety of open source explicit, and in marking the emerging sub-genres, namely co-opetition, protestware, and appropriate technology, where new empirical work is most needed.