Pattern no. 4 — “Minimising the surface of the common”: Architecture · Rule
Translated from the French, which remains the reference text. This translation is under review.
Pattern n°4 · Stratification (primary), Metabolism, Thirdness · tensions T1, T10, T3 · Legitimacy · volets I.2, II.2 · ↑ navigation
Schema F4.1 · Context of pattern no. 4
Tension(s) at work: Innovation and standardisation; simplicity and complexity; autonomy and connectedness. This is the mother fiche of the language: it answers the question all the others presuppose resolved — what goes into the waist of the hourglass, and above all, what is kept out of it?
Context: An ecosystem is designing its common layer. Around the table, each stakeholder arrives with a legitimate demand: its use case to cover, its data field to standardise, its indispensable feature. Each addition, taken in isolation, is reasonable; their sum is a certain catastrophe. For the surface of the common is not neutral: each element added to the waist will have to be implemented by everyone, maintained forever, and secured everywhere. The standardisation committee is a thickening machine — that is its natural slope, and no one escapes it without an instituted discipline.
Forces at play: On one side, inclusive gravity: covering more use cases seems to serve adoption, and refusing a member's demand is politically costly — “no” has no constituency. On the other, the arithmetic of adoption: each additional page of specification raises the barrier to entry, excludes small implementers, multiplies the attack surface, slows evolution — and, beyond a threshold, turns the commons into its opposite: a protocol grown so thick that one must be a platform to implement it recreates the platform. Conversely, a commons that is too thin no longer holds the network together: interoperability dies of what is not specified. The median compromise — “optional modules” — fails consistently: the option the dominant player implements becomes everyone else's de facto obligation.
The pattern: Admit into the waist only what passes three tests — and institute the power to say no. The necessity test: does the network break without this element? If it can interoperate without it, the element goes to the bells. The stability test: will this element vary across territories, uses, periods? What varies is heavy — it is territorialised; only the invariant enters the waist (this is the protocol side of the cosmo-local test of “Arriving first at the periphery”, n°1). The simplexity test: can an ordinary developer implement it in days, not months? Implementation experience is not a documentation veneer, it is an admission criterion — Aadhaar's authentication API is the ideal type: it answers yes or no, one bit, and that bit has carried a billion identities; all the complexity (biometrics, deduplication, security) is buried beneath the surface, never exposed by it. Complexity hides under the bonnet; it is not handed out to implementers. And because the three tests do not apply themselves, the discipline is instituted: someone must have the mandate to refuse — saying no is the first function of the specification's guardian, and a sound roadmap is steered as much by subtraction as by addition.
Operator(s) mobilised: Primary: stratification — this fiche is its design rule. Support: thirdness — the power to say no cannot belong to an interested member; it falls to the guardian (fiche n°9), the only actor whose mission is the health of the waist rather than its capture. Support: metabolism — each line of specification is a perpetual maintenance liability; the surface of the common is its cost of ownership, and whoever thickens the waist signs a debt in everyone's name (fiche n°10).
The invariant under test: Minimalism's own risk is exclusion by omission: what the waist does not specify, the powerful specify de facto — “optional” accessibility does not exist for those who depend on it, data protection left “to implementers” means the dominant player's default, and the apparently neutral protocol then encodes, implicitly, the preferences of the strongest. Associated criterion of legitimacy, which ties this fiche to n°6: the surface is minimised, the invariant never — the values coded into techno-legal artefacts (consent, rights, inclusion) are the only non-negotiable thickness of the waist. The tension between the two fiches is assumed and productive: “Minimising the surface” pulls towards less, “Coding the law” imposes a floor of more — and it is their friction, arbitrated case by case by the guardian, that draws the right waist. A fit hourglass is narrow and ballasted.
Schema F4.2 · Minimising the surface of the common
Cases of embodiment: Indian: India Stack is the historical demonstration of the pattern — APIs of archetypal minimalism (authentication in one bit, payment in one virtual address), an explicit doctrine of infrastructural frugality (“the rails, nothing but the rails”), and the well-known result: tens of thousands of integrations, because any team of three developers could implement in a week. European: the Dataspace Protocol is the successful internal counter-model — a deliberately narrowed scope (contract negotiation and transfer, nothing else: no storage, no application catalogue, no business layer), where the first generation of European initiatives had erred through the opposite ambition. The comparative lesson: the thin waist is not a renunciation of ambition — it is its vehicle; India Stack does more than anyone precisely because its waist does less.
Counter-example: Gaia-X, in its first form. The initiative wanted to be everything at once — service catalogue, labelling system, federation architecture, policy framework, technical standard, and industrial alternative to the American cloud: a maximal surface, embracing infrastructure, standard and market in a single gesture. The result was adoption inversely proportional to ambition: industrial players waited, developers did not know where to begin, competitors laughed — and each found, somewhere in the immense scope, its own particular reason not to come. The later refocusing — towards a narrowed trust framework and decoupled specifications — is precisely a movement of surface reduction, and it must be read as such: not a failure, but the learning of the pattern in real conditions, of which today's Data Spaces are the chastened heirs. The theorem: an infrastructure that wants to do everything gives everyone a reason not to come — one does not adopt a cathedral, one visits it.
Signs of misapplication: The specification grows at every committee meeting. The number of pages grows faster than the number of independent implementations. “Optional” features are required in practice by the dominant implementation. The reference implementation time is counted in months. The standard specifies choices that belong to the bells (storage, interface, business logic). No one around the table has the explicit mandate to refuse. The roadmap contains only additions — no deprecation, ever. The team of three developers cannot get started in a week (the India Stack test). Two symptoms warrant an alert; three, a diagnosis.
Related patterns: This fiche is upstream of all the others: “The whole in every part” (n°5 — what spreads everywhere must be small), “Running code first” (n°2 — what is delivered quickly must be simple), “Paying for the first tank” (n°3) and “Arriving first at the periphery” (n°1 — only the adoptable can be bootstrapped), “Those who prosper maintain” (n°10 — the surface is the cost). Normative hinge: “Coding the law” (n°6 — the non-negotiable thickness, the assumed counterweight to minimalism).