Common stakes
Translated from the French, which remains the reference text. This translation is under review.
Introduction: Embracing complexity
As Edgar Morin very rightly observes in La Méthode: “The choice […] is between mourning and the search for a method able to articulate what is separated and to reconnect what is disjoined. […] Today, our historical need is to find a method that detects [and instruments] the linkages, articulations, solidarities, implications, imbrications, interdependencies, complexities.”
At the close of this report, we believe we have brought to light a single search, conducted on two continents, for systems capable of articulating what the previous cycle had disjoined — the social and the technical, the distributed and the federated, the symbolic and the neuronal, the national and the multilateral. Taken together, these dynamics combine, in their processes as in their purposes, principles that one would at first consider antagonistic. On closer inspection, however, they articulate dialogically, in the sense Morin gives the term: they remain antagonistic and competing, yet prove complementary — without any synthesis ever having to dissolve the tension that makes them productive.
Yet this embraced complexity, so often held against the third way as its congenital flaw — too many forms of governance, too many standards, too many stakeholders, where the hegemonies advance as one — is precisely what cybernetics invites us to read differently. The law of requisite variety, formulated by W. Ross Ashby, states that a regulating system must have at least as much internal variety as the system it claims to regulate. Applied to our object, it turns the criticism into a project: faced with the complexity of the IRL and digital worlds — billions of users, thousands of jurisdictions, an irreducible diversity of languages, cultures and economic models — the two hegemonies, American and Chinese, proceed by reducing variety. The American platforms absorb it by centralising data; the Chinese model absorbs it by centralising authority. The third way is the only one that claims to absorb complexity by embracing it: through federation rather than fusion, through subsidiarity rather than uniformity, through the minimal protocol rather than the total platform. Its internal variety is not its weakness; it is its condition of viability and its differentiating advantage — on the sole condition, to which we shall return, that its engineering is mastered.
7.1Eleven tensions
The report met these conjugations of antagonists at every stage. They are worth inventorying.
- Innovation and standardisation: they do not stand opposed, contrary to an intuition widespread among many digital actors; it is standardisation that enables innovation. Such is the concept, the strategy and the reality of India Stack's hourglass, and more broadly of the Indian approach; going one step further, standardisation must itself be considered an innovation — an organisational one. This articulation is also at the heart of the European approach (Data Strategy, Data Spaces).
- Bottom-up and top-down: the point is not to choose, but to articulate. Bottom-up gathers the greatest possible amount of information (collective intelligence) and fosters ownership; it is necessary, productive and salutary, but it produces divergence. Top-down then intervenes as arbiter, to settle, to standardise, and immediately to give rise to bottom-up again: innovation on rails, feedback and evolvability. The two logics articulate sequentially, recursively. Mechanisms exist to nest one within the other — this is the very essence of representative democracy.
- Autonomy and connectedness: sovereign strategic autonomy at the micro level (the actors — Data Spaces, ONDC) and the macro level (countries, continental blocs), and common rails — standards, interoperability — at those same two levels. Decentralisation and federation.
- Uncoordinated and coordinated action: NGI brought gems to light in a deliberately uncoordinated way (NextGraph, among others); the Open Internet Stack now aims to coordinate and integrate that ecosystem. Emergence first, integration next;
- Diversity and unity: Morin's hologrammatic principle finds here its most literal embodiment. The protocol is hologrammatic in the strict sense: the whole — the rules of the network — is entirely present in each of the parts — each implementation — and it is because it is that the parts can form a network without a centre. The concept of protocol, at the heart of the Indian paradigm as of the European one, is the technical answer to a philosophical question.
- Local and global: and finally cosmo-localism itself — mutualise the light, territorialise the heavy — as the central issue of the third way (MOSIP being its most accomplished example).
- Cooperation and competition: This tension is the very mantra this report has seen formulated, in almost identical terms, from one continent to the other: cooperation and competition. Cooperate on rails, compete on services; public rails, private innovation. Cooperating on the common layer to make competition possible — and fruitful — on the application layers: it is the most operative conjugation of antagonists in our whole corpus, the one whose economic evidence is the most massive (UPI, ONDC), and the one that most sharply distinguishes the third way from platformised winner-takes-all as from state dirigisme.
Schema 7.1 · Openness comes from underneath
- Long term and agility: holding the vision — the third way — over the long term, while continuously integrating new data (the irruption of generative AI), opportunities (the Indo-European synergy) and constraints (technical debt, socio-technical complexity). The resolution here lies not in space but in time: it is a matter of cycles, of sequences, of breathing.
- Technical and legal: computer code and legal code, long thought of as two alien normative orders, and which this report has seen merge into an artefact of a new kind — the techno-legal: legal code through computer code. The Rulebooks and ODRL policies of the Data Spaces, and DEPA's cryptographic consent, are its first industrial embodiments.
- Simplicity and complexity: the power of simplicity (India Stack and its minimalist APIs) and the power of complexity (Confidential Clean Rooms, Data Spaces) —, whose reconciliation has a name, simplexity (Alain Berthoz): not simplification, which amputates, but the art of making complexity practicable — what software engineering calls complexity hiding.
- Public, private and citizen: a triad rather than a dyad: all have a role, and none suffices. The public regulates and enables; the public, the private and citizens innovate, co-produce, fund and use.
What simplicity asks in practice is reported by five architecture fiches in chapter 5: what Aadhaar, eSign, DigiLocker, UPI and the account aggregators require of a person and what they keep, set beside their European counterparts. Architecture fiche A · IdentityArchitecture fiche B · SignatureArchitecture fiche C · DocumentsArchitecture fiche D · PaymentArchitecture fiche E · Data
7.2Five operators, one invariant
Can we go beyond the mere inventory of these tensions? We believe so. For in observing how the ecosystems studied resolve them in practice, a small number of ways of doing things recur with striking regularity, which we name operators. Three of them, synchronic, resolve tensions of structure; two others, diachronic, which the challenges presented in the next section make necessary, resolve tensions of dynamics.
Schema 7.2 · Five operators, two kinds of tension
Tensions of structure
the three synchronic operators
Stratification
minimise the surface of the commons
Alternation
never freeze the hourglass — turn it over
Thirdness
the commons needs a guardian that does not own it
Tensions of dynamics
the two diachronic operators
Bootstrapping
a commons does not bootstrap itself — someone has to pay for the first tank of fuel
Metabolism
those who prosper on the rails maintain the rails
Five operators
- Stratification: resolution in logical space. Never do the systems studied seek the midpoint between two poles; they separate them into layers. Unity is pushed towards the thinnest possible layer — the protocol — and variety released everywhere else. This is subsidiarity applied to technical architecture: India Stack, Beckn, the Dataspace Protocol — and TCP/IP before them — all obey the same design rule: minimise the surface of the commons. A commons that is too thick becomes a platform again; a commons that is too thin no longer holds the network together. The whole art lies in specifying the waist of the hourglass.
- Alternation: resolution in time. When two logics cannot coexist simultaneously, they are sequenced: exploratory divergence, then arbitral convergence, then fresh divergence on the basis thus stabilised. NGI brings things to light, the OIS integrates; bottom-up proposes, top-down settles, feedback corrects; this is the cycle of the IETF's RFCs as of well-designed public policy. The associated rule: never freeze the hourglass — turn it over. A standard that ceases to be re-questioned becomes a fossil; an exploration that is never arbitrated becomes entropy.
- Thirdness: resolution by a third party. When two poles can neither merge nor ignore one another, the systems studied institute a third party that holds the tension without belonging to either. Institutional third party: the neutral steward, decoupled from states as from markets — MOSIP hosted at IIIT-B, Samyog, the Eclipse or NLnet foundations. Artefactual third party: the techno-legal object itself, which is law and code without being reducible to either — cryptographic consent, the executable Rulebook. Organisational third party: the not-for-profit data intermediary of the Data Governance Act, the Indian Section 8 Companies. The associated rule, which chapter 6 illustrated and which pattern no. 9 will formalise: the commons needs a guardian that does not own it.
These three operators (synchronic) say how to design the system. But an infrastructure system is not only an architecture — it is born, it survives, it wins or it loses. The hardest challenges presented in the next section — seizing the momentum, winning the protocol war, crossing the valley of death — are not tensions of structure but trials of dynamics: the system against its environment, and against time. Yet the corpus, here again, shows that the ecosystems that succeed always answer them in the same way, through a fourth and a fifth operator (diachronic) that complete the grammar:
Schema 7.3 · Alternation: the cycle does not return to where it began
exploratory divergence
- then
arbitral convergence
- then
fresh divergence
the basis thus stabilised
what arbitration has settled, and what the next exploration starts from
never freeze the hourglass — turn it over
- Bootstrapping: resolution in dynamics. Every infrastructure commons faces the chicken-and-egg problem: no users without services, no services without users, and network effects reward the first to arrive. The cases studied never rely on spontaneous generation: one actor takes it upon itself to ignite the network. The state as reference user (Aadhaar backed by social programmes, public procurement made a lever by the EOS); the enrolled champion (the banks brought on board UPI); the offer made to the world (MOSIP freely proposed to the global South). The associated rule: a commons does not bootstrap itself — someone has to pay for the first tank of fuel.
- Metabolism: resolution in economy. A commons is an organism: it needs a regular flow of resources, not drip feeds. Commons do not die first of war; they die of the free rider and of unfunded maintenance — the precariousness of project-based funding, the valley of death between grant and market, which this report has documented, are its symptoms. The forms of response observed: the levy taken on the rails themselves, the mutualised maintenance instrument (at the heart of the EOS), the contribution of beneficiaries proportionate to the benefits — the principle of congruence that Elinor Ostrom identified as vital to any commons. The associated rule: those who prosper on the rails maintain the rails.
One invariant: the Values / Identity / Legitimacy triptych
Is the grammar then complete? Cybernetics offers a test: Stafford Beer's Viable System Model — a direct heir of Ashby — enumerates the functions without which no system remains viable. Our five operators correspond, to a first approximation, to the functions of operation and cohesion (stratification, thirdness), of adaptation (alternation, bootstrapping) and of supply (metabolism). There remains the function Beer places at the summit of every viable system: identity — the instance that says who we are, and what we will not do.
We suggest that identity should not be made a sixth operator, but an invariant. And that it must moreover be distinguished from two neighbouring notions:
- Values — consent, inclusion, dignity, sovereignty are the content, that is, what, in a system designed throughout to vary, must precisely not vary. It is in this sense that we call values the invariant.
- Identity is the function that carries it: the organ where the system keeps memory of what it is — and that function has, in the architecture of the third way, a precise location: the waist of the hourglass, where techno-legal artefacts encode values into the protocol itself.
- Legitimacy, finally, is neither the content nor the function: it is the proof — the always provisional verdict, rendered from outside and continuously by those the system claims to serve. The Aadhaar paradox showed it: an infrastructure that claims inclusion will be judged on the exclusion it produces or avoids — and no operator can render that judgement in its place. The operators say how to operate; values say in the name of what; legitimacy says whether it held.
The attentive reader will have noted that the word “invariant” has already appeared in this chapter, to designate the waist of the hourglass — the technical invariant that seven tensions led us to encounter. The collision is not fortuitous; it is perhaps the central discovery of this report: the techno-legal is precisely the gesture by which the normative invariant inscribes itself in the technical invariant — values in the protocol and, through the hologrammatic principle, in each of its implementations. Five operators, one invariant: the grammar is complete. The system, for its part, must never be — a closed system is, as thermodynamics has taught us, a dead system. It is even precisely because it remains open to every wind of variety that it needs a fixed point: one needs an invariant only once one has decided that everything else might change.
Schema 7.4 · The content, the function that carries it, and the proof rendered from outside
Identity
the function that carries it: the organ where the system keeps memory of what it is
Values
the content: that which, in a system designed throughout to vary, must precisely not vary
- consent
- inclusion
- dignity
- sovereignty of the person
inscribed in
the waist of the hourglass, where techno-legal artefacts encode the values in the protocol itself
Legitimacy
neither the content nor the function: the proof — the always provisional verdict that the invariant has held in the real world
rendered from outside and continuously by those the system claims to serve
7.3Seven challenges
If dialogic tensions are the material of the next cycle, challenges put that material to the test. Each is the shadow cast by a tension identified above. This is precisely why the preceding grammar is more than a speculative exercise: each challenge can be linked to the operator or operators through which it can be addressed. We organise them below around the distinction that now structures our grammar: structural challenges, which the synchronic operators can address; dynamic challenges, which call on the diachronic operators and are the hardest, because they pit the system against its environment and against time; and finally the test of the invariant, which is never resolved once and for all.
Structural challenges
- Engineering simplexity. The first challenge is the shadow of the interface tension: the third way bets on embracing complexity rather than reducing it, but that complexity must remain manageable in practice. A Data Space whose onboarding takes months of work by lawyers and integrators, or a Confidential Clean Room that only ten actors worldwide know how to operate, will never become a commons: it will become a cathedral. The lesson from the corpus comes from India: India Stack's strength lies in the minimalism of its APIs. All the complexity — biometrics, interbank operations, consent — sits beneath interfaces that any developer can master within days. The operator at work is stratification, developed into a discipline of its own: making developer experience and integration experience primary design criteria, alongside security and compliance. The challenge remains largely unresolved in Europe, where the dense layers of rules surrounding Data Spaces and their cumbersome connectors threaten to break the hourglass.
- Bringing the archipelago together. The second challenge is the shadow of the tension among actors and the uncoordinated/coordinated pair: the proliferation of initiatives — NGI, OIS, Eurocommons, STF, EDIC Digital Commons, MOSIP and others — demonstrates the movement's vitality while threatening to fragment it. How can they work together and converge without merging, which would recreate the very platform they seek to escape? The answer lies in the most elegant property of our grammar: it is fractal. All that is needed — if we dare put it that way — is to apply to the ecosystem of initiatives the operators each already applies to its own work. Stratification: a minimal common invariant (MUST) between initiatives, as proposed by the DSIF initiative for Data Spaces and by Digital New Deal, with nothing superfluous. Thirdness: a third party that holds the whole together without owning it — a role neither the Commission nor any Member State can fulfil alone, and for which no institution currently takes responsibility. Alternation: organised cycles of convergence, beginning with the Euro-Indian workshops proposed in the conclusion of this report. The challenge is therefore one of institutional engineering, rather than conceptual design, and it is urgent.
Dynamic challenges
- Seizing the momentum: the window at the edge. “History is written by the victors” — and in digital infrastructure, the victors are almost always the first to arrive: network effects turn a lead into a rent, and that rent into an irreversible position. Tackled head-on, this challenge would already be lost: the centralised cloud layer is entrenched, and a twenty-year gap cannot be closed. But the premise of this report shifts the front line: if advances in models, local-first approaches and neuro-symbolic methods allow most of what centralised AI does today to run at the edge, then there is a layer of the stack — local computing and inference close to where it is needed — where history has yet to be written. The third way's goal is therefore to arrive first at the edge. The operators at work are alternation — delivering before the framework of rules is perfect — and bootstrapping — concentrating public procurement and funding instruments on this window while it remains open. This framing gives momentum a time-bound, actionable target. It makes decentralised computing the chosen terrain of the contest.
- Winning the protocol war — to avoid becoming the next OSI. The third challenge is the shadow of the search for a common invariant. History helps illuminate it. In the 1980s and 1990s, two visions of network interconnection competed: OSI, the multilateral, institutional and comprehensive standard backed by standards bodies and governments, particularly in Europe; and TCP/IP, backed by an engineering community whose motto consisted of five words: rough consensus and running code. OSI had legitimacy, completeness and public mandates on its side; TCP/IP had working code. We know which prevailed. With its committees, consortia and governance frameworks, the third way faces a structural risk of becoming the next OSI. The response has three parts. Alternation: deliver reference implementations before taking standards to the IETF, W3C or ISO. Bootstrapping: win through adoption rather than committees — the MOSIP strategy, which gained adoption in twenty-five countries without a single standardisation battle. Finally, defensive thirdness: licences and governance arrangements that make it legally costly to pursue the embrace, extend, extinguish strategy through which dominant powers absorb open standards.
- Horizontal integration: supply — and demand. The fourth challenge is the shadow of coopetition at three scales: micro — enabling competing software vendors to cooperate; meso — bringing together continental blocs that each aspire to lead; and macro — offering the world a unified proposition. The equation is the same at every scale: cooperating on the rails requires each participant to give up ownership of them, a concession that neither commercial interests nor national pride readily make. The need to work on the supply side is established: without credible, integrated alternatives that are ready for production and maintained, the third way remains a matter of words. This is the significance of the transition from NGI to the Open Internet Stack. But the assessment of Europe in this report also requires attention to demand: once an offering exists, it still has to be bought. Legislation and billions in investment have not been enough to redirect public procurement, change established habits or overcome switching costs. This is where bootstrapping takes on its full political significance: the state as a reference user is an indispensable condition, rather than simply one industrial-policy option. Europe has much to teach India about safeguards; in return, it must learn from India about adoption.
- Crossing the valley of death: sustainability. The fifth challenge is the shadow of metabolism, and perhaps the most insidious, because it emerges through silent exhaustion rather than a crisis. Commons succumb first to free-riding and unfunded maintenance, rather than attacks by their opponents. The corpus documents the symptoms: precarious project-based funding, the valley of death between an R&D grant and the market, and critical components maintained by tiny teams. The EOS Maintenance Instrument signals recognition of the problem. A lasting arrangement still has to be devised — a levy on the rails, with beneficiaries contributing in proportion to the benefits they receive — to turn a drip feed into a metabolism. The principle is simple to state and politically difficult to establish: those who prosper on the rails maintain the rails. None of the ecosystems studied has fully resolved this challenge; the first to do so will gain a decisive advantage.
- Resisting capture. The sixth challenge is the defensive counterpart of the third, Winning the protocol war, and overlooking it would be naive: the duopoly will not stand by as a third way takes shape. Its instruments are familiar and already in use: legal extraterritoriality — the CLOUD Act, which strips data localisation of its substance on infrastructure subject to US law; capture of standards bodies, where the dominant powers are strongly represented; open-washing, through which hyperscalers embrace open source and, more recently, sovereignty in order to neutralise them; and the capture of talent, against which no European or Indian instrument can currently compete. The response draws on defensive thirdness — stewards whose governance prevents takeovers, licences designed as armour — and on techno-legal mechanisms that encode sovereignty in the artefact as well as the contract. It also requires a clear recognition that the third way is engaged in a struggle for power.
Schema 7.5 · Openness and verifiability