Status: bootstrap / draft institutional repository. Organization governance, Founding Stewardship, Membership, the constitutional framework and the Contributor License Agreement are non-operative unless an exact adoption record expressly states otherwise.
This repository is the organization-level control plane for Exergism Commons (EC): constitutional rules, institutional authority, founder-led bootstrap, membership, contributor rights, intellectual-property provenance, stewardship, delegations, decision classes, domain custody and other cross-project policies.
Project-specific philosophical, analytical, evidentiary, legal and technical authority remains in the repository that owns that layer. Nothing here silently changes an Exergism release, ECL Bundle, ECL Schedule, ECL-PL patent grant, Funding record or other immutable project artifact.
Mission / Constitution / adopted organization policy
↓
Founding phase + Membership + explicit roles
↓
versioned decision rule
↓
machine-readable governance state
↓
SHACL / deterministic integrity checks
↓
GovernanceDecision / Delegation
↓
domain implementations
├── Funding
├── id.exergism.org
├── repositories
├── contributor rights / CLA
└── future registry
The machine layer enforces a reviewable subset of adopted rules. It cannot manufacture legal or institutional authority.
CONSTITUTION.md— proposed constitutional framework and Mission Lock.FOUNDING-STEWARDSHIP.md— founder-led bootstrap, Founding Steward powers/limits, Mission Veto, F0→F1→F2 transition and succession.MEMBERSHIP.md— one-person-one-vote membership, admission, Candidate period, voting seasoning, inactivity, suspension and anti-Sybil controls.ARCHITECTURE.md— cross-project authority boundaries.architecture/DOMAIN-AND-URI-ARCHITECTURE.md—exergism.org, persistent identifiers and infrastructure custody.IP-POLICY.md— cross-project intellectual-property and inbound-rights policy.CONTRIBUTING.md— how to propose organization-level governance changes.OPEN-KNOWLEDGE-POLICY.md— proposed Public Knowledge, Source Form and anti-enclosure policy.
spec/MACHINE-READABLE-GOVERNANCE.md— authority hierarchy, exact threshold arithmetic, Membership/Founding projections and validation boundary.policy/governance-status.json— top-level institutional/adoption state. Bootstrap is fail-closed withoperative: false.policy/decision-rules.json— exact rational rules forOrdinaryApproval,QualifiedApproval,ConstitutionalAmendmentandMissionLockedAmendment.policy/membership-status.json— draft repository-safe Member Registry projection.policy/founding-stewardship.json— draft Founding Steward, Mission Lock and maturity-phase projection.policy/delegations.json— explicit delegation registry; currently empty/non-operative.policy/open-knowledge-status.json— closed-world machine projection of the Open Knowledge/anti-enclosure contract.ontology/commons.ttl— shared cross-project EC primitives athttps://id.exergism.org/commons#.ontology/governance.ttl— organization governance vocabulary athttps://id.exergism.org/governance#.ontology/commons-context.jsonld— shared Commons JSON-LD context.ontology/governance-context.jsonld— Governance JSON-LD context, with shared Commons primitives imported through theecprefix.ontology/governance-shapes.ttl— structural SHACL constraints.tools/validate.py— canonical complete fail-closed integrity entrypoint. It installs strict parsing, lifecycle, temporal, release-history, authority, Open Knowledge and CLA/Schedule gates before returning PASS.tools/validate_governance.py— internal core validation library. Running/importing this module alone is not the supported activation verdict.tools/test_governance_mutation_guards.py— adversarial mutation matrix for recurring bug classes such as subset weakening, paired-source edits and mutable taxonomy redefinition.
cla/CLA-1.0-DRAFT.md— proposed individual Contributor License Agreement.cla/ENTITY-CLA-1.0-DRAFT.md— proposed entity/corporate Contributor License Agreement.cla/PROJECT-SCHEDULE.md— repository-by-repository scope and outbound constraints.cla/PROCESS.md— proposed acceptance, verification, records and exception process.cla/PRIVACY-AND-RECORDS.md— privacy, data minimization, retention and agreement-record architecture.cla/LEGACY-CONTRIBUTIONS.md— treatment of contributions made before CLA adoption.cla/AI-ASSISTED-CONTRIBUTIONS.md— provenance rules for AI-assisted work.cla/DESIGN-RATIONALE.md— design rationale and rejected alternatives.cla/FAQ.md— contributor-facing explanation.cla/ADOPTION.md— activation gates.policy/cla-status.yaml— machine-readable CLA adoption state.policy/covered-projects.yaml— machine-readable CLA project scope projection.
The complete validation command is:
python tools/validate.pyCI also runs python tools/test_governance_mutation_guards.py. The distinction is intentional: the first validates the actual repository state and historical authority chain, while the second mutates whole classes of rights-sensitive fields to prove the validator fails closed under adversarial combinations.
For rights-sensitive schema versions, EC uses a closed-world contract: required fields are compared as exact sets/mappings, not non-empty, all true, minimum-length or subset checks. Likewise, agreement between two mutable representations is insufficient for authority-sensitive semantics. The Schedule and its YAML projection are both checked against an independent semantic invariant set, so changing both sides to the same weakened value is still a failure.
Temporal authority is also historical rather than current-state-only. Operative release history is validated from the permanent release-1 founding anchor forward; a later amendment cannot retroactively repair a missing founding Open Knowledge binding, backdate a newly introduced policy, or authorize itself with proposed replacement rules.
Exergism
└─ canonical philosophy + formal analytical model
ECL
└─ ECL evidence/governance + copyright/software-rights legal artifacts
ECL-PL
└─ separate optional patent-licensing architecture
Funding
└─ funding strategy, records and executable funding-policy subset
id.exergism.org
└─ persistent identifier resolution, not semantic authority
www.exergism.org
└─ public presentation, not canonical authority
Exergism Commons governance
└─ organization-level institutional authority and shared constraints
No arrow, repository membership, funding relationship, semantic edge or technical permission transfers authority automatically.
EC currently depends materially on its founder. This repository makes that concentration explicit instead of pretending that a mature democratic institution already exists.
The proposed initial Founding Steward is Daniel Molinero Lucas.
The principle is:
strong stewardship without ownership.
During F0 the Founding Steward has strong strategic/bootstrap authority, but founder status creates no automatic salary, grant percentage, Endowment claim, Treasury ownership or IP ownership.
The Founding Steward also has a narrow negative Mission Veto during the Founding Period. It can block a proposal that violates the protected Mission Lock or attempts a protected irreversible action through the wrong approval path. It cannot enact an alternative decision or override conflict rules for the founder's own private benefit.
Founder executive power is designed to recede through evidence-based maturity phases:
F0 founder-led bootstrap
↓
F1 early institution
↓
F2 distributed institution
At F2 the founder is expected to cease having general executive primacy and retain, if still applicable, a narrower Mission Guardian function.
Membership governs Exergism Commons, not an arbitrary organization that happens to inherit its assets and name.
The Mission Lock protects EC's institutional identity: stewardship around Exergism, project-authority separation, non-ownership membership, no governance-for-money, contributor-rights attribution, persistent-ID integrity, contestability, anti-capture and anti-enclosure of established EC Public Knowledge.
It does not freeze every philosophical claim, roadmap item or implementation decision.
A MissionLockedAmendment is deliberately harder than an ordinary constitutional amendment: during the Founding Period it requires at least 3/4 quorum, 90% approval, two successful votes at least 60 days apart, independent review and Founding Steward/Mission Guardian consent.
Voting membership is for natural persons and follows:
one person = one Member = one vote.
Founder status, funding, employment, GitHub permissions, commit count or donations do not increase voting weight.
The draft lifecycle is:
participant
↓
Candidate (>=30 days)
↓
Active Member
↓
class-specific voting seasoning
Default minimum Active-Membership age before voting:
- Ordinary Approval: 30 days;
- Qualified Approval: 90 days;
- Constitutional Amendment: 90 days;
- Mission-Locked Amendment: 180 days.
This reduces capture through mass admission immediately before consequential votes.
The proposed bootstrap record designates Daniel Molinero Lucas as the initial Member for formation purposes, but that projection remains operative_membership: false until a competent institutional adoption creates the actual Member Registry.
Member status does not grant bank, DNS, repository-admin, CLA-record or other operational access; roles and delegations are separate.
Thresholds are encoded as exact integer ratios rather than rounded decimals:
- Ordinary Approval — quorum >1/2 of non-conflicted eligible voters; more votes for than against.
- Qualified Approval — quorum ≥2/3; approval ≥2/3.
- Constitutional Amendment — quorum ≥2/3; approval ≥3/4.
- Mission-Locked Amendment — quorum ≥3/4; approval ≥9/10 plus the additional Mission Lock process.
Domain policies may require stricter thresholds but must not silently weaken a named organization-level approval class.
The conflict model expressly prevents a compensation beneficiary from substantively voting on their own compensation. Founder status does not bypass recusal. Funding cannot purchase governance rights.
Organization governance defines shared authority concepts; domain repositories implement them.
Funding may require ecg:QualifiedApproval for a concentration or Endowment decision while remaining authoritative for the funding facts and records that trigger that requirement. Likewise, id.exergism.org may consume the organization rule that permanent domain/identifier-authority transfer requires Qualified Approval while remaining only the resolver implementation.
Persistent vocabularies:
- shared Commons vocabulary:
https://id.exergism.org/commons# - shared Commons ontology:
https://id.exergism.org/ontology/commons - Governance vocabulary:
https://id.exergism.org/governance# - Governance ontology:
https://id.exergism.org/ontology/governance
Minting a term does not make a policy operative.
The proposed model is license-in, rights-retained:
- contributors keep the copyright they own;
- EC receives limited review/archive rights while an unaccepted proposal is under review;
- the full perpetual project grant vests only for an Accepted Contribution;
- outbound relicensing is constrained rather than unlimited;
- the CLA itself grants no patent rights;
- repository administration does not create ownership; and
- institutional succession cannot silently rewrite earlier grants.
The same acceptance boundary now applies to both the Individual and Entity CLA drafts.
The constitutional, Founding Stewardship, Membership, Open Knowledge and CLA packages are intentionally non-operative at bootstrap.
tools/validate.py is the supported complete fail-closed verdict. Merely flipping an operative boolean must fail unless the associated legal entity/adoption metadata, Member Registry, founder assignment, conflict/records/treasury controls, legal review, Open Knowledge release history and aligned policy flags are present. Lower-level validator modules exist as implementation components and must not be treated as equivalent standalone activation verdicts.
Before organization governance can become operative, EC must resolve at least the legal entity/form, governing law, initial membership/voter registry, valid adoption mechanism, records/privacy controls, treasury authority, Founding Steward assignment/succession compatibility, an adopted rights-reviewed Open Knowledge implementation and appropriate independent legal review. The CLA has additional activation gates in cla/ADOPTION.md.
A merge of draft architecture into Git is not ratification, incorporation, signature, delegation or legal advice.