Pattern no. 2 — “Running code first”: Strategy · Practical stake
Translated from the French, which remains the reference text. This translation is under review.
Pattern n°2 · Alternation (primary), Bootstrapping, Thirdness · tensions T1, T8 · Legitimacy · volets I.2, I.3 · ↑ navigation
Schema F2.1 · Context of pattern no. 2
Tension(s) at work: Innovation and standardisation; long term and agility; and, implicitly, the quest for the common invariant — for this pattern answers the question: how does a standard become the invariant of the network, rather than a document?
Context: A multilateral ecosystem wants to establish a common protocol — the waist of its hourglass. Several institutional paths are open to it: formal standardisation (ISO, committees, diplomatic consensus) or de facto standardisation (adopted implementations, engineers' consensus). The ecosystem is plural by construction: many stakeholders, many jurisdictions, strong demands for procedural legitimacy — everything pushes it towards the formal path.
Forces at play: On one side, legitimacy: a multilateral standard must be able to claim a fair, documented process open to all — otherwise it reproduces the capture it claims to fight. On the other, speed and technical truth: a standard exists only if it is implemented, and the market does not wait for committees; while one standardises, the adversary ships, and network effects turn its lead into irreversibility. The median compromise — standardising “a little faster”, implementing “a little in parallel” — regularly fails: it produces lukewarm standards that no one implements with conviction and implementations that no one recognises as a reference.
The pattern: Deliver the reference implementation first; take the standard to the formal bodies afterwards — and only what usage has validated. The formal standardisation process is not abolished: it is repositioned downstream, as an instrument for the diplomatic consolidation of what adoption has already settled. The canonical sequence: (1) an open source reference implementation, operated by a neutral steward; (2) real adoption by actors who commit their production to it; (3) the specification extracted from usage, stabilised through short cycles; (4) finally, formal standardisation (ISO, CEN) as an act of international recognition — not as a birth certificate.
Operator(s) mobilised: Primary: alternation — the pattern is purely a matter of sequence; it does not change the ingredients, it changes their order. Support: bootstrapping — the reference implementation is the “first tank” that someone must pay for, and early adoption is won, not waited for. Defensive support: thirdness — the reference implementation must be held by a guardian who does not own it (foundation, SRO), otherwise the de facto standard becomes the de facto property of its first implementer.
The invariant under test: The pattern has a risk of its own: the engineers' consensus that replaces diplomatic consensus can exclude silently — those without the means to implement no longer have a say, and rough consensus has historically been the consensus of the best-resourced. Associated criterion of legitimacy: the governance of the reference implementation must guarantee an effective right to implement (documentation, tooling, support — MOSIP's frugal engineering is the model), failing which the “open” standard rebuilds a barrier to entry.
Schema F2.2 · Running code first
Cases of embodiment: Indian: UPI never waited for an international standard — NPCI delivered the system, the banks adopted it, volume did the rest; formal recognition and the international interconnection accords came after the fact, in the position of petitioners. The same sequence for Beckn: protocol published, implementations delivered (ONDC, Namma Yatri), and it is only then that Europe takes an interest (Beckn-LD, Open Commerce Bridge). European: the Dataspace Protocol follows — a notable fact for Europe — the right sequence: an open source reference implementation (the Eclipse connector, under a neutral foundation), real deployments in the Data Spaces, then the move to ISO standardisation as consolidation. It is the internal counter-model of the European normative tradition, and it must be protected as such.
Counter-example: OSI versus TCP/IP, 1980s-1990s — the founding failure, and the mirror held up. OSI had everything: the exemplary multilateral process, the support of governments (up to the GOSIP public procurement mandates), conceptual completeness in seven layers, the legitimacy of the standards bodies. TCP/IP had running code, free implementations in every university, and a motto: rough consensus and running code. The perfect standard died in the drawers; the imperfect protocol conquered the world — then treated itself, after victory, to all the formal standardisation it wanted. The lesson is not that standardisation is futile: it is that arriving first, it kills; arriving second, it crowns. The third way, with its consortia and its governance frameworks, is structurally exposed to replaying OSI. This fiche exists to prevent it.
Signs of misapplication: The specification is larger than the code that implements it. The working group counts more lawyers than developers. The roadmap promises standardisation before the first production deployment. The “reference” implementation is held by an actor that sells services on top of it. Revision cycles are counted in years. Two symptoms warrant an alert; three, a diagnosis.
Related patterns: Upstream: “Minimising the surface of the common” (n°4 — what should be standardised?). Downstream: “The guardian who does not own” (n°9 — who holds the reference implementation?) and “Those who prosper maintain” (n°10 — who funds its maintenance?).