Pattern no. 8 — “Cooperate on the rails, compete on the services”: Institutionalisation · Boundary
Translated from the French, which remains the reference text. This translation is under review.
Pattern n°8 · Thirdness (primary), Metabolism, Stratification · tensions T11, T7 · Legitimacy · volets I.2, II.1 · ↑ navigation
Schema F8.1 · Context of pattern no. 8
Tension(s) at work: Cooperation and competition — the reigning dyad of the corpus, the one whose mantra was uttered, almost word for word, on two continents (public rails, private innovation; cooperate on rails, compete on services); public, private and citizen. Implicitly, the challenge of horizontal integration, which arises identically at three scales — micro (competing software publishers sharing a protocol), meso (continental blocs sharing a layer of infrastructure), macro (a common offer proposed to the world): at each scale, making competitors cooperate without merging them, and without any of them capturing what they share. The question this fiche examines is not whether to cooperate on the rails — the corpus has settled that — but exactly where the boundary between the rail and the service runs, and who guards it.
Context: Competing actors have an interest in sharing a layer — identity, payment, discovery, trust, data transport — because the fragmentation of that layer would destroy value for everyone. Above it, they want to compete, because that is where value is created and differentiated. The principle is simple; its implementation is the site of every ruse. For the boundary between rail and service is never given by the technology: it is decided, and every actor has an interest in moving it — the rail operator upwards (adding services to its natural monopoly), the dominant service provider downwards (absorbing the rail to make it its moat).
Forces at play: On one side, the operator's vertical gravity: whoever holds the rail sees every flow pass, knows every actor, and always ends up wondering why it should leave the upper margins to others — self-preferencing is the natural slope of every infrastructure. On the other, the dominant player's inverse vertical gravity: whoever wins on the services ends up wanting to own the rail, to turn its lead into a perpetual rent — vertical integration is the natural slope of every winner. The median compromise — the “open platform” operated by one of the competitors, coopetition proclaimed without instituted separation — fails invariably: a rail owned by a competitor is not a rail, it is a toll.
The pattern: Draw the boundary explicitly, found it on a criterion, and structurally forbid its crossing in either direction. Three constitutive moves. The sharing criterion: a rail is what competition would destroy the value of (interoperability, trust, identity — a second payment standard is not competition, it is fragmentation); a service is what competition creates the value of (experience, price, innovation in use). The rule fits in one line: cooperate where competition destroys, compete where it creates. Structural separation: the rail operator does not compete on the services — NPCI favours no application, not even BHIM, its own; ONDC sells nothing; and the European Data Governance Act has made this a legal obligation for data intermediation service providers (DISPs), who are bound not to exploit in any way the data they carry and to isolate the intermediation in a distinct legal person —; reciprocally, no service provider holds the rail. The boundary is written into the statutes and, now, into the law — not into intentions. Equal access: the same interfaces, the same conditions, the same published tariffs for all competitors — non-discrimination is not a promise, it is a verifiable property of the rail. And because the boundary is a decision, it has a governance: it is revised, by the third party that guards it, when technology or the market shift the line between what fragments and what differentiates — yesterday payment was a service, today it is a rail; tomorrow discovery, or inference.
Operator(s) mobilised: Primary: thirdness — the boundary can be guarded neither by the rail operator nor by the competitors on the service; it requires a third party that is neither one nor the other (fiche n°9), and whose first function is precisely to say this is rail, this is service. Support: stratification — the rail/service boundary is the very line that separates the waist of the hourglass from its bulbs; this fiche is the economic face of “Minimising the surface of the common” (n°4): what enters the rail ceases to be a field of competition. Support: metabolism — who pays for the rail is the twin question of who owns it (fiche n°10): a rail funded by the competitors but owned by none is the stable form; a rail funded by the public and exploited by private actors is the unstable form, see below.
The invariant under test: The risk proper to the pattern is a silent inversion: the rhetoric of rails can serve to socialise costs and privatise benefits — the public builds and maintains the infrastructure, private actors harvest the upper margins, and users pay twice (through tax, then through the service). The Indian critique of the Aadhaar/UPI pair bears exactly there: public infrastructure, fintech profits. Criterion of legitimacy: cooperation on the rails must produce a benefit verifiable for those it serves, not only for those who compete on top of it — a measured fall in prices, measured inclusion, and the publication of those measurements. Second risk: competition on the services can degenerate into a race to the bottom on the protection of persons (collection, consent, dark patterns) if the rail does not carry the invariant — hence the junction with “Coding the law” (n°6): values are encoded in the rail, precisely because it is the only layer where competition cannot erode them. A legitimate rail is one on which no one can compete against those it serves.
Cases of embodiment: Indian: UPI is the planetary demonstration of the pattern — a single rail, operated by a not-for-profit utility, and above it a fierce competition between applications that has made India the world's first market for instant payment. But the case also teaches that the boundary demands active guarding: the concentration of services on two dominant applications led the regulator to consider a market-share cap — coopetition does not maintain itself, it is regulated. ONDC pushes the pattern one notch further by unbundling the service itself: buyer applications and seller applications kept separate, so that no marketplace can reconstitute itself on the rail. European: the historical precedent belongs to Europe, and it must be claimed — GSM: competing operators and manufacturers cooperating on a single standard, then competing on networks and handsets; the continent's greatest digital industrial success is an instituted coopetition. The Data Spaces are its conscious heir — Catena-X brings rival carmakers together on a common data rail, with the mantra on its banner — and the eIDAS wallet reproduces the separation: a public identity rail, private services competing above it.
Counter-example: Two tolls teach the pattern by the negative, each in one direction of crossing. The card networks — Visa, Mastercard — are a rail owned by its participants and funded by a proportional levy on every transaction: the interchange fee, one to three per cent of world commerce, collected by a duopoly that no one can bypass; the rail has become the most lucrative of rents because it belongs to those who compete on it. The App Store is the inverse crossing: a service provider — Apple, publisher of its own applications — owns the distribution rail of its competitors, sets its conditions at thirty per cent and exempts itself; self-preferencing so structural that it took a continental law (the DMA) to name it. The theorem, twofold: whoever owns the rail taxes the service; whoever provides the service and owns the rail eliminates its competitors. The pattern exists so that neither sentence becomes true.
Signs of misapplication: The rail operator launches “an offer” on the upper layers. The conditions of access to the rail are not published — or differ from one actor to another. Access to the interfaces requires a non-disclosure agreement or a partnership. Two actors concentrate most of the services and no body has a mandate to respond. The rail's fees are neither at cost nor publicly justified. The public funds the infrastructure and no measurement of the public benefit is published. The rail/service boundary has not been revised since the foundation. Two symptoms are a warning; three, a diagnosis.
Related patterns: Upstream: “Minimising the surface of the common” (n°4 — the waist of the hourglass is the rail; this fiche is its economic face) and “Coding the law” (n°6 — the invariant is encoded in the rail because it is the only layer outside competition). Downstream: “The guardian who does not own” (n°9 — the third party that guards the boundary) and “Those who prosper maintain” (n°10 — who funds the rail that no one owns). Further on: “The armour of the common” (n°11 — defending the rail against its absorption by a dominant service).