Pattern no. 9 — “The guardian who does not own”: Institutionalisation · Governance
Translated from the French, which remains the reference text. This translation is under review.
Pattern n°9 · Thirdness (primary), Metabolism, Stratification · tensions T11, T3, T7 · Legitimacy · volets I.3, II.2 · ↑ navigation
Schema F9.1 · Context of pattern no. 9
Tension(s) at work: Autonomy and connectedness; public, private and citizen; and, implicitly, coopetition — for this pattern answers the question every pooling leaves open: if the rails belong to no one, who holds them? A commons without a guardian is prey; a commons whose guardian is its owner is no longer a commons.
Schema F9.2 · Public, private and citizen
Context: An ecosystem has pooled a critical layer — protocol, reference implementation, cryptographic keys, registry, brand. This layer must be operated over time: maintained, governed, defended, arbitrated. Yet none of the natural candidates will do. The State? No foreign State adopts code owned by another — and the domestic State brings the risk of surveillance and of political reversal. A company? It will end up monetising the bottleneck; that is its function. A consortium of the dominant actors? It will close the door behind it. The context is all the more acute when the commons is multilateral: for a Euro-Indian layer, neither Brussels nor New Delhi can hold the keys — the neutrality constraint is squared.
Forces at play: On one side, the need for a strong operator: orphaned commons die, and the diffuse “community” signs no contracts, answers to no court, keeps no security on-call rota. On the other, the certainty of capture: any holder of the critical assets — code repository, brand, keys, the right to modify the specification — ends up tempted, bought out or compelled to monetise them. The median compromise — “balanced governance within the leading company”, “an advisory committee attached to the ministry” — fails systematically: power follows the assets, not the organisation charts.
The pattern: Institute a third party whose statutes make ownership impossible — and entrust it with the critical assets, not merely with their facilitation. Thirdness is not decreed by good will; it is engineered through four cumulative locks. The mission lock: a not-for-profit structure whose corporate purpose is the commons itself, with an inalienability clause — on dissolution, the assets pass to another commons, never to a member. The asset lock: the guardian holds in its own right what commands the network — the reference repository, the brand, the specification, and where relevant the keys (Samyog's Key Management Service is its pure form: the separation of control guarantees that no participant can access the data alone). The funding lock: plural resources, capped per contributor — a guardian 80% funded by a single actor is a subsidiary that does not know it. The governance lock: no college — States, industry, community — holds a majority on its own, and the guardians' mandates are revocable by those they guard.
Operator(s) mobilised: Primary: thirdness — this pattern is its constitutive rule. Support: stratification — entrust the guardian with the waist of the hourglass only, never the bells: a steward that also operates services competes with its own principals and loses its neutrality through the mere play of its interests. Support: metabolism — the four locks hold only if the guardian is funded on a permanent footing (see fiche n°10); a poor guardian sells, a precarious guardian gives way.
The invariant under test: The pattern's own risk is subtle: neutrality can degenerate into unaccountability. A guardian that belongs to no one may answer to no one — and the “foundation” then becomes a screen (foundation-washing is an industry). Associated criterion of legitimacy: neutrality is measured by accountability, not by independence — transparency of accounts and decisions, avenues of recourse for members and end users alike, and the effective revocability of those in charge. A legitimate guardian is one that the guarded can judge — and replace.
Schema F9.3 · The guardian who does not own
Cases of embodiment: Indian: MOSIP is the textbook case of geopolitical neutrality by construction — decoupled from the Indian State, hosted by an academic institution (IIIT-Bangalore), funded by philanthropic foundations precisely because no foreign State would have adopted code owned by New Delhi; its neutrality is not a virtue, it is its condition of commercial existence. Samyog embodies the asset lock pushed to the extreme: a not-for-profit self-regulatory organisation whose sole reason for being is to hold what neither the State (risk of surveillance) nor the private sector (risk of monetisation) should hold — the keys. The Section 8 Companies (ONDC, NPCI) show the hybrid version: the State gives the impetus, the private sector funds the bootstrapping, the not-for-profit status locks out the rent. European: the Eclipse Foundation plays the role of neutral home for the reference connector of the Data Spaces — an asset that no member company holds; the foundations mandated by NGI's cascade funding (NLnet) and the DGA's data intermediary status, bound to commercial neutrality by law, complete the picture: Europe has the particularity of having begun to codify thirdness in law.
Counter-example: MySQL, a free database that became infrastructure of the web — but whose assets (brand, repository, copyright) belonged to a company: bought by Sun, then Sun by Oracle, the commons found itself the property of its main competitor. The community survived only through secession (the MariaDB fork) — that is, by paying the maximum price: rebuilding everything, the brand included. The same film played out for OpenOffice, swallowed in the same acquisition, and saved by the creation a posteriori of the guardian that should have existed a priori — The Document Foundation, born in haste to host LibreOffice. The lesson stands as a theorem: when the guardian does not exist at the outset, the community always ends up creating it — but after the crisis, and at the price of a schism. The OpenAI case supplies its contemporary, inverted variant: a not-for-profit guardian did exist, but without asset or funding locks — and the structure mutated under the pressure of the capital it was meant to contain. Statutes without locks are wishes.
Signs of misapplication: The brand or the reference repository is in the name of a member company. A single funder exceeds a third of the resources. The guardian sells services on the layers it arbitrates. Governance minutes are not public. Dissolution is possible with no devolution clause. Those in charge cannot be removed, or are appointed by a single college. The word “foundation” is in the name, but the commons' developers are employed by a single actor. Two symptoms warrant an alert; three, a diagnosis.
Related patterns: Upstream: “Running code first” (n°2 — the reference implementation that has to be housed somewhere) and “Cooperate on the rails, compete on the services” (n°8 — the boundary the guardian guards). Downstream: “Those who prosper maintain” (n°10 — the metabolism without which the locks give way) and “The armour of the common” (n°11 — the defensive tooling the guardian wields).