STStratosGovernance diagnostic

Open source governance and considerations

Stratos governance diagnostic. Version 0.1.

This document covers the open-source dimensions of the Stratos project: licensing, contributor agreements, governance structure, IP boundaries, code of conduct, legal and geopolitical risk, and the practical considerations of running a politically sensitive open-source project as an independent researcher.

Open-sourcing a governance-analysis project is different from open-sourcing a typical library. The risks split into four categories: licence compatibility (data and code), IP boundary management (especially around methodology that has independent value), exposure (governments and political actors push back when measured), and reputational liability (methodology errors become public errors). All four are addressable. None of them are reasons to keep the project closed.

This document is intended to be read by potential contributors, by users assessing the project's credibility, by lawyers if the project ever requires legal review, and by the maintainer as a reference for decisions that will recur.

1. Licensing strategy

Stratos releases components under different licences depending on what each component is and how it is intended to be used.

1.1 Code licence: MIT

The code is licensed under the MIT licence.

Why MIT. MIT is permissive. It allows commercial use, modification, distribution, and private use, subject only to including the original licence notice. It is the most widely understood open-source licence and imposes minimum friction on contributors and users.

Why not AGPL. AGPL would prevent proprietary forks from operating Stratos-derived dashboards as a service without contributing improvements back. This is theoretically appealing as a defensive measure. In practice, AGPL is a contributor deterrent. The project's value comes from broad adoption and contribution, not from preventing competitive forks. The same logic applies to GPL.

Why not Apache 2.0. Apache 2.0 adds explicit patent grant language, which is valuable in some contexts. For Stratos, the methodology has independent IP handling (see section 4), and the code itself is mostly application logic without patentable inventions. MIT's simplicity is preferable.

What MIT does not protect. MIT does not prevent commercial forks. It does not require attribution beyond the licence notice. It does not require contribution back. These are accepted as design choices, not problems.

1.2 Documentation licence: CC-BY 4.0

Documentation, including the methodology specification, this governance document, and the seed documents, is licensed under Creative Commons Attribution 4.0 International.

Why CC-BY. Documentation is intended to be read, reproduced, translated, and incorporated into other works (academic citations, journalistic coverage, civil society reports). CC-BY is the standard licence for this use case. Attribution is required, which preserves credit, but no other restrictions apply.

Why not CC-BY-SA. ShareAlike would require derivative works to carry the same licence. For documentation, this would limit incorporation into mixed-licence works (academic papers, news articles, etc.) and reduce the documentation's reach. The cost outweighs the benefit.

1.3 Derived data licence: ODbL with attribution

Derived data products (the computed dimensional scores, the country profiles, the trajectory data) are licensed under the Open Database License (ODbL) where the underlying source licences permit.

Why ODbL. ODbL is the standard licence for open data. It permits reuse, modification, and redistribution, with attribution and ShareAlike on the data itself. It is compatible with most source licences in the data sources audit.

Where source licences impose restrictions (CC-BY-NC, CC-BY-ND, non-commercial restrictions), the affected derived data is either:

  1. Isolated into a separately licensed module clearly marked with the source restrictions, or
  2. Excluded from the published derived data products

This isolation is essential for downstream users who need to know whether commercial use is permitted.

1.4 Source data licensing

Source data is governed by the licensing terms of each source. Stratos does not redistribute source data wholesale. It uses source data as input to its computations and provides clear pointers to the original sources.

For sources where licensing terms permit redistribution of cleaned and harmonised versions of the source data, Stratos may publish such versions in the data/processed/ directory of the repository. For sources where licensing terms restrict redistribution, only the derived outputs are published, and users are pointed to the source for the underlying data.

The DATA_LICENCES.md document tracks the licensing status of each source and the corresponding handling.

2. Contributor licence and agreement

The project uses a lightweight Developer Certificate of Origin (DCO) model for contributions, not a full Contributor Licence Agreement (CLA).

2.1 The DCO approach

Contributors sign off on their commits with the DCO statement (the standard Signed-off-by: line). The DCO asserts that the contributor has the right to submit the contribution under the project's licence and that they understand the contribution is public.

Why DCO and not CLA. A CLA assigns more extensive rights to the project maintainer or a sponsoring organisation. It is appropriate for projects with substantial commercial considerations or institutional backing. The DCO is sufficient for a project with the structure Stratos has: independent maintainer, MIT-licensed code, no commercial entity behind the project.

If the project ever attracts commercial backing or transitions to a foundation governance model, a CLA may become appropriate. The DCO is the right choice for the current stage.

2.2 What contributions are accepted

The contribution scope is documented in CONTRIBUTING.md. Acceptable contributions include:

  • Code: bug fixes, feature implementations, refactoring
  • Documentation: corrections, clarifications, translations, expansions
  • Data sources: integration of new sources following the audit principles
  • Methodology: substantive methodology proposals through the issue tracker
  • Country-expert review: validation and correction of diagnostic narratives
  • Design: frontend design improvements, accessibility improvements

Contributions that are not accepted:

  • Political advocacy unrelated to methodology
  • Personal commentary on specific officials or political parties
  • Promotional content for external products or services
  • Speculative methodology changes without substantive analysis

2.3 Maintainer decision authority

The founding maintainer retains final authority on all contributions through Phase 1 of the roadmap. This is the benevolent-dictator model and it is the appropriate governance for an early-stage independent project.

After Phase 1, governance transitions to a documented process described in section 5.

3. Code of conduct

The project adopts the Contributor Covenant version 2.1 as its code of conduct. The full text is included in CODE_OF_CONDUCT.md.

3.1 Scope

The code of conduct applies to all project spaces: the GitHub repository, issue and pull request discussions, project communications channels, and any in-person or virtual events organised under the project's name.

3.2 Enforcement

The founding maintainer is the initial enforcement authority. Reports go to a designated email address documented in the code of conduct.

Enforcement follows the Contributor Covenant's documented procedures: investigation, response, appropriate action (warning, temporary ban, permanent ban). Enforcement decisions and reasoning are documented privately.

As the project grows, enforcement authority may transition to a small committee with rotating membership. This transition is documented if and when it happens.

3.3 The political-sensitivity question

Governance analysis is politically charged. The project will attract contributors with strong political views. The code of conduct provides the foundation for managing this productively.

Three additional norms supplement the standard code of conduct:

Methodology over politics. Substantive methodology disagreements are welcome and important. Discussions that pursue political positions through methodology arguments are redirected.

Comment on the work, not the contributor. Critique of methodology, code, or documentation is welcome. Critique of the contributor as a person is not.

Disagreement is acceptable; harassment is not. Strong disagreement, expressed civilly, is part of productive discourse. Personal attacks, sustained hostility, or targeting of contributors based on their identity is not.

4. IP boundary management

Stratos has external methodology dependencies that need clear IP separation. The most important is Probability Cloud Regression.

4.1 Probability Cloud Regression

PCR is the intellectual property of its developer (Soliman, 2026). It is used in Stratos but is not part of Stratos.

The IP boundary is implemented as follows:

  • The PCR methodology is documented and (where the developer chooses) published in academic venues separately from Stratos
  • The PCR implementation lives in a separate repository with its own licence
  • Stratos imports PCR as a dependency
  • Stratos contributors do not contribute to PCR through the Stratos repository
  • Stratos can use PCR. PCR is not bound by the Stratos licence.

This separation matters for two reasons. First, PCR has applications beyond governance analysis, and its independent IP allows the developer to pursue those applications. Second, the academic credit for PCR remains with the developer regardless of how Stratos evolves.

The boundary is not unusual. Many open-source projects depend on libraries with different IP arrangements. The novelty here is the explicit documentation of the boundary, which is appropriate given the methodological significance of PCR.

4.2 Other methodology contributions

The project welcomes substantive methodology contributions. Where a contributor makes a substantive methodological contribution (a novel scoring approach, a new conditioning technique, an improved uncertainty quantification method), the contribution is:

  • Credited to the contributor in the methodology documentation
  • Subject to the same documentation standards as the rest of the methodology
  • Released under the project's licences once accepted

Substantive methodology contributions that the contributor wishes to retain as independent IP can be implemented as dependencies (following the PCR pattern) rather than as direct contributions. The maintainer is open to either model.

4.3 Academic credit

Stratos is structured so that academic credit can flow appropriately to contributors and dependencies. Three mechanisms:

  1. Citation file: CITATION.cff in the repository root specifies the recommended citation for the project as a whole
  2. Methodology citations: the methodology specification cites contributing methodologies (PCR, V-Dem, etc.) appropriately
  3. Contributor recognition: substantial contributors are recognised in documentation and acknowledgements

Academic users citing Stratos in their work follow the citation file. The project's reputation accumulates over time through such citation, which benefits both the project and its contributors.

5. Project governance

Governance evolves as the project grows. Three phases are anticipated.

5.1 Phase A: benevolent dictator (current)

The founding maintainer is the sole decision authority on all aspects of the project: methodology, code, scope, contributions, releases. This is appropriate for an early-stage project where coherence and direction are more important than distributed decision-making.

Decisions of consequence are documented:

  • Methodology decisions in METHODOLOGY_CHANGELOG.md
  • Roadmap decisions in ROADMAP_DECISIONS.md
  • Technical decisions in TECHNICAL_DECISIONS.md

The maintainer commits to documenting reasoning for substantive decisions even when the decision is unilateral. This is part of the open-development practice.

5.2 Phase B: small maintainer team

When the project has at least three regular contributors with sustained involvement, governance transitions to a small maintainer team. The transition criteria:

  • At least three contributors have been actively involved for at least six months
  • At least two contributors have demonstrated capability to make substantive merge decisions
  • The contributors are willing to take on ongoing maintainer responsibility

The transition is documented in GOVERNANCE.md with explicit changes to decision authority.

In Phase B, the founding maintainer retains authority over methodology direction and overall project scope. Day-to-day decisions (merges, releases, issue triage) are distributed across the maintainer team.

5.3 Phase C: documented governance with broader community

When the project has substantial contributor base (a dozen or more regular contributors), governance may transition to a more formal model with documented decision procedures, lazy consensus on routine decisions, and explicit voting on substantive changes.

This transition is anticipated but not committed. It depends on the project growing in a way that makes formal governance valuable rather than burdensome.

5.4 Funding and sponsorship governance

The project does not currently accept funding. If funding becomes available, three principles apply.

Funding source transparency. All funding sources are disclosed publicly. There is no anonymous funding.

Funding conditions transparency. Any conditions attached to funding are disclosed publicly.

Funding does not buy methodology influence. Funders cannot direct methodology decisions. They can fund work, but the work is judged by the same standards as any other contribution.

These principles protect the project's credibility, which is more valuable than any specific funding opportunity.

Open-source governance analysis carries risks that ordinary open-source projects do not face. The risks are addressable through careful design.

6.1 Government pushback

Governments do not always appreciate being measured. The risk profile varies dramatically by regime cluster.

Cluster A and B democracies typically tolerate critique. Methodology criticism may come from government-aligned scholars or media but does not threaten the project's operation.

Cluster C and D countries sometimes apply diplomatic pressure to organisations that publish unfavourable analysis. For an independent project without institutional backing, this pressure is less effective than against a university or NGO, but it can take other forms: attempts to influence platform hosting, complaints to funders if funders exist, public attacks on the maintainer.

Cluster E (closed autocracies) sometimes apply more aggressive measures: blocking access in their territory, harassment of contributors, attempts to influence platforms used by the project, occasionally legal action in jurisdictions where it might succeed.

Mitigation strategies:

  • Methodology is fully transparent. Critique is invited and responded to substantively.
  • Diagnostic narratives are framed as descriptive analysis under explicit methodology, not as judgement.
  • Confidence bands are visible. Strong claims require strong evidence.
  • Contributors can use pseudonyms if they have reason to believe their identity creates risk.
  • The maintainer is based in a jurisdiction (Australia) with strong protections for academic and analytical work.
  • The project does not accept funding from any government, which protects against accusations of partisan motivation.

Country-level analysis (rather than commentary on specific officials) generally does not create defamation exposure. The project stays at country and government-as-entity level.

Three practices reduce legal exposure:

Methodology, not commentary. Diagnostic narratives describe what the methodology produces. They do not editorialise about individual policymakers, parties, or officials.

Sources cited. Every claim is traceable to a source. Where the project's methodology produces a strong claim, the methodology is the source, and the methodology is fully public.

Disclaimers visible. The dashboard carries clear disclaimers about its purpose (research and education), limitations (methodology and data), and the appropriate use cases (analysis and discussion, not investment or political decision-making).

Where the project receives a takedown demand, the response process is:

  1. Acknowledge receipt promptly
  2. Review the substance with an experienced contributor familiar with the relevant content
  3. Respond substantively, defending content that is methodologically sound
  4. Correct content that is methodologically unsound regardless of who pointed out the problem
  5. Document the exchange publicly when appropriate

The project does not capitulate to legally weak demands but does correct genuine errors.

6.3 Platform risk

The project depends on third-party platforms: GitHub for code, Vercel for hosting, npm and PyPI for dependencies, domain registrars, DNS providers.

Platform risk is real but not unique to this project. Mitigation is standard:

  • Code is in Git, which is portable across hosting platforms
  • The application is a standard Next.js app, deployable to any equivalent host
  • Domain registration is with a registrar that supports transfer
  • DNS is portable
  • The deployment infrastructure is documented so a migration could be executed if needed

The probability of platform action against the project is low but non-zero. The mitigation is not paranoid but is documented.

6.4 Contributor safety

Contributors based in some jurisdictions may face risk for participating in a project that analyses their own government critically. This is a real consideration.

Three practices apply:

Pseudonyms accepted. Contributors can participate under pseudonyms. The project does not require legal identity for contribution.

No required disclosure of location. Contributors are not required to disclose their location publicly.

Personal communications protected. The maintainer does not disclose contributor identities or locations to third parties without contributor consent.

These practices are documented in CONTRIBUTING.md and SECURITY.md.

7. Liability disclaimers

The dashboard and its documentation carry visible disclaimers covering three areas.

Methodology limitations. The dashboard is a research and educational tool. Its outputs reflect methodology choices that are documented but contestable. Users are encouraged to engage with the methodology, not just the outputs.

Not professional advice. The dashboard is not investment advice, sovereign-risk analysis, political analysis, policy recommendation, or any other form of professional advice. Users who need professional advice should consult professionals.

Use at own risk. The MIT licence's standard warranty disclaimer applies. The project is provided as-is. Users assume the risk of their reliance on its outputs.

The disclaimers are visible on the dashboard, in the documentation, and in the licence files. They are not buried.

8. AI use disclosure

The project is built by an independent researcher using AI-assisted coding tools (Claude Code, Codex, GitHub Copilot). This is disclosed publicly because it is honest, because it is part of the maintainer's broader practice of AI-use transparency, and because users have a legitimate interest in knowing how the dashboard is constructed.

The disclosure is positioned as a strength of the methodology rather than as a vulnerability. Three reasons:

Methodology is human-designed. The substantive methodology decisions (regime classification scheme, conditioning approach, scoring weights, uncertainty quantification) are made by the maintainer. The AI tools accelerate implementation, not methodology.

Implementation is human-supervised. Every line of code is reviewed by the maintainer before merge. AI suggestions that do not meet the project's quality bar are rejected.

Diagnostics are template-driven, not free-generation. The diagnostic narratives use parameterised templates written by the maintainer. The dashboard does not use free-form LLM generation for its public outputs.

The AI use disclosure is on the methodology page, in the contributor documentation, and (eventually) in any academic publication associated with the project.

9. Sustainability

Open-source projects fail more often than they succeed. The maintenance burden grows over time while contributor enthusiasm typically declines. Three sustainability practices apply.

Scope discipline. The roadmap has explicit kill switches. Features are added when they justify their maintenance cost. Features are removed when they no longer do.

Documentation as memory. Decision records, changelogs, and roadmap documentation reduce the cost of contributor turnover. A new maintainer can read the project's history and continue work without losing context.

Sustainable maintenance pattern. The maintainer commits to sustainable work patterns, not heroic effort. The project is structured so that the maintenance burden is bounded.

The project is open about the possibility that it might be archived if sustainability cannot be maintained. Archiving with explicit reasoning is preferable to slow neglect.

10. Closing notes for potential contributors

Two practical notes for contributors thinking about getting involved.

Start small. The project welcomes substantial contributions, but the best way to begin is with a small contribution: a documentation correction, a bug fix, a translation of a single page. Small contributions let both the contributor and the maintainer learn how to work together.

Engage on methodology if that is your interest. The methodology is the heart of the project. If you have substantive methodology expertise, the most valuable place to contribute is through the methodology discussion on the issue tracker. Code contributions are welcome but methodology contributions move the project further.

The project is built for the long term. Contributors who engage substantively over time become part of how it evolves.