An Empirical Study of Blockchain Repositories in GitHub

1 INTRODUCTION

• Prior GitHub studies focused on specific or small Blockchain-repository samples, leaving the ecosystem’s diverse states and user interactions unexamined.Blockchain solutions had also become more diverse than cryptocurrency offerings alone.Source: §1 · p. 1
• The study manually categorizes 3,664 repositories as tools or applications, then examines GitHub metrics to characterize user engagement and interactions.Applications are divided into cryptocurrency and non-cryptocurrency categories.Source: §1 · pp. 1–2
• Organizations contribute more than individual users; tools show greater collaboration than applications, commits dominate interaction types, and issues exceed commits and pull requests in interactions.Crypto applications are more popular, while crypto and non-crypto applications have mostly similar development activity.Source: §1 · p. 2
• Many internal developers face restrictions on merging changes, and the authors recommend measures to improve their autonomy.The study also identifies implications for developers, vendors, and researchers, including awareness of ecosystem trends and quality-assurance tools.Source: §1 · p. 2
• The authors suggest Blockchain vendors improve repository autonomy, developers learn current ecosystem trends, and researchers develop quality-assurance tools for software and user interactions.Source: §1 · p. 2

2.1 Project Selection

• Because GitHub lacks a standard Blockchain-repository label and keyword searches can miss or falsely include projects, the study searched 86 terms across repository names, descriptions, and topics.The terms cover cryptocurrency platforms, permissioned Blockchains, and popular Blockchain-related concepts.Source: §2.1 · p. 2, Figure 1 · p. 2

2.2 Project Filtering

• Filtering candidates by size, stars, recent activity, data availability, and software-project content reduced the pool to about 5,200 repositories, followed by manual review yielding 3,664.Manual review removed irrelevant and non-source-code repositories.Source: §2.2 · pp. 2–3

3 EMPIRICAL STUDY

• RQ1 examines repository popularity and activity through repository growth and user activity.Source: §3 · p. 3
• RQ2 concerns the distribution of different users in repositories, while RQ3 examines how users collaborate.Source: §3 · p. 3
• RQ4 asks how autonomous internal users are, specifically whether they can progress without much help from external users.Source: §3 · p. 3

3.1.1 Approach.

• Repositories were classified as Tool, Crypto application, or Non-Crypto application; keyword automation categorized about 25% of the dataset, and coders manually checked every repository.Tools offer a stand-alone API/SDK, while applications address specific use-case scenarios.Source: §3.1.1 · p. 3

3.1.2 Results.

• Ethereum (31.7%) and Bitcoin (15.9%) projects occupy the largest reported shares of the studied repositories.The platform ranking places Ethereum first and Bitcoin second.Source: Figure 2 · p. 4, §3.1.2 · p. 4
• Tool repositories have more commits, issues, and pull requests than applications, while crypto and non-crypto application activity distributions do not differ (p > 0.05).The study reports no significant activity-distribution difference between crypto and non-crypto applications for any of the three activity measures.Source: §3.1.2 · p. 4
• Kruskal-Wallis tests found popularity and activity distributions differed across repository types (p < 0.001); Dunn’s tests with Bonferroni adjustment found Tool and Application repositories did not differ significantly in stars or forks.Source: §3.1.2 · p. 4
• Popularity medians are similar across the three categories, although quartile differences are greater between Tools and Applications Others.Figure 3 presents the distributions of stars and forks by category.Source: Figure 3 · p. 4, §3.1.2 · p. 4

3.2.1 Approach.

• The study classifies repository users as internal or external by summarizing their contributions, including who processes issues and pull requests.Internal users who significantly participate in processing others’ issues or pull requests are marked as maintainers.Source: §3.2.1 · p. 4

3.2.2 Results.

• Tool repositories have higher numbers of both internal and external users than application repositories.Figure 5 shows the distributions separately for internal and external users.Source: §3.2.2 · p. 5, Figure 5 · p. 5
• The most internally and externally active repositories differ by category, including bitcoin (614 internal users) and go-ethereum (4,402 external users) among tools.For crypto applications, gekko has the most internal users (199), while metamask-extension has the most external users (3,013); for non-crypto applications, the corresponding leaders are condenser (82) and monero-gui (560).Source: §3.2.2 · p. 5
• User distributions differ significantly across repository types, but internal-user distributions for crypto and non-crypto applications do not differ significantly.External-user distributions for tools and crypto applications also do not differ significantly (p > 0.05).Source: §3.2.2 · p. 5
• Organizations own 58.6% of the studied Blockchain repositories, and tools have higher median numbers of both internal and external users than applications.Ethereum repositories are mostly tools, while IBM repositories are mostly applications.Source: §3.2.2 · p. 5

3.3 RQ3. How do the different users collaborate in the Blockchain repositories?

• The study measures collaboration through contribution, maintenance, process, review, and discussion interactions among repository users.Contribution tracks commit authors and committers; maintenance includes users approving or commenting on artifacts such as pull requests.Source: §3.3.1 · p. 5, §3.3.2 · pp. 5–6

3.3.1 Approach.

• Median users per artifact are higher in tools than applications: commits 15.5, issues 10, and pull requests 7 in tools.The corresponding medians are 10, 8, and 4 for crypto applications and 12, 9, and 7 for non-crypto applications.Source: §3.3.2 · p. 5

3.3.2 Results.

• Code contribution is most prevalent in tools and non-crypto applications, while discussion is most prevalent in crypto applications.Issue maintenance and issue processing exceed their pull-request counterparts in all repository types; review is higher on pull requests than issues for tools and non-crypto applications.Source: §3.3.2 · p. 6
• Figures 6–8 present collaborative interactions per artifact for tools, crypto applications, and non-crypto applications, respectively.The plotted scales differ across the graphs.Source: Figure 6 · p. 6, Figure 7 · p. 6, Figure 8 · p. 6
• Contribution interactions on commits outnumber almost all other interaction types, and issues have more interactions overall than commits or pull requests.Collaboration is higher in tools than applications, including maintenance interactions.Source: §3.3.2 · p. 6

3.4 RQ4. How autonomous are the internal users in the Blockchain repositories?

• Repository autonomy is calculated from the percentages of users classified as maintainers, autonomous contributors, or dependent contributors.The study examines the distribution of these three user types across Blockchain repository categories.Source: §3.4.1 · p. 6, §3.4.2 · p. 6

3.4.1 Approach.

• The proportion of maintainers is higher in crypto applications than in tools and non-crypto applications.Figure 9 compares the proportions of maintainers, autonomous contributors, and dependent contributors across repository categories.Source: §3.4.2 · p. 6, Figure 9 · p. 7

3.4.2 Results.

• Table 5 is identified as the source for the user types used in the autonomy calculation.The supplied table caption does not provide the definitions or category values.Source: Table 5 · p. 7

4.1 Blockchain Vs Non-Blockchain Repositories

• Blockchain repositories have more commits, issues, and pull requests than the sampled non-Blockchain repositories, alongside higher internal and external user counts.Organizations own 58.6% of Blockchain repositories versus 24.9% of non-Blockchain repositories; Figure 10 compares collaborative interactions per artifact.Source: §4.1 · p. 7, Figure 10 · p. 8
• Median collaboration frequencies are higher in Blockchain repositories across contribution, maintenance, process, review, and discussion interactions.Maintainers’ proportions are similar, but autonomous users are 33% in Blockchain repositories versus 40% in non-Blockchain repositories, while dependent contributors are 33% versus 25%.Source: §4.1 · p. 8

4.2 The Case of Archived Blockchain Repos

• Archived repositories comprise 9.1% of Blockchain repositories and 8.6% of non-Blockchain repositories, with a Blockchain archival upswing in 2017–2018.Repository creation also rose in 2017–2018 for both groups, more prominently for Blockchain repositories; Figure 11 tracks archived and newly created projects.Source: §4.2 · p. 8, Figure 11 · p. 8
• In a sample of 200 archived repositories, archival reasons were unreported for more than half; obsolete use cases and new-repository migration were more common in Blockchain repositories.The sample comprised 100 repositories of each type, selected for the reasons analysis.Source: §4.2 · pp. 8–9

4.3 Implications of Findings

• The discussion suggests repository owners consider less restrictive environments because Blockchain repositories show lower user autonomy than non-Blockchain repositories.It also advises developers to expect more activity and collaboration in tool repositories and possible review cycles when applying changes.Source: §4.1 · p. 8, §4.3 · p. 9
• The authors suggest researchers consider tools for repository autonomy and automatic documentation, noting the attention users give to issues.They also point career enthusiasts to repository growth trends as a source of ecosystem context.Source: §4.3 · p. 9

• Three human coders were used to mitigate author bias in manually labeling Blockchain repositories.Source: §5 · p. 9
• Cohen’s κ showed substantial-to-perfect agreement between the first two coders across two iterations.Source: §5 · p. 9
• The study used topics and project descriptions alongside established GitHub-repository analysis practices to address methodological errors.Source: §5 · p. 9
• The 86-keyword sampling approach was necessary because GitHub has no universal Blockchain identifier and topic tags are incomplete.Many identified repositories had no topic tag; the keyword search returned 802K repositories, which the authors considered sufficient to find a good number of Blockchain repositories.Source: §5 · p. 9

• Earlier empirical studies focused on particular domains or relatively small cryptocurrency samples, while this study examines Blockchain repositories in the wild.Related work also included literature reviews, taxonomic surveys, and studies of domain-specific applications.Source: §6 · p. 9
• Prior technical work addressed vulnerabilities, scalability, and automated smart-contract vulnerability detection.Examples include Vandal, ContractFuzzer, and TEETHER.Source: §6 · p. 9
• The study analyzes a large GitHub sample and examines both developer collaboration and user autonomy across the Blockchain OSS ecosystem.Source: §6 · pp. 9–10

• 3,664 GitHub repositories form the empirical study of a rapidly growing Blockchain ecosystem that peaked in 2016–2018.The first Blockchain project appeared in 2010, and the peak period coincided with adoption in domains beyond cryptocurrency.Source: §7 · p. 10
• Large organizations are more present in Blockchain repositories and own multiple repositories on GitHub.Source: §7 · p. 10
• Users show less autonomy in Blockchain repositories; future work will investigate its causes and develop guidelines and tools.Source: §7 · p. 10