Memo 08 — International Governance: Norm Production Without a Sovereign

Series: The Landscape of Normative Systems (LNS) Series ID: LNS Memo number: 08 Primary JD questions: Q1 (sources of normative authority above the state), Q2 (six-part per-source analysis), Q6 (the EU and international organizations as norm-generating systems). Secondary: Q5 (feedback loops), Q7 (Politics/Law/Power/Legitimacy/Trust relations under anarchy), Q8 (what must stay above the kernel). Status of D0: Immutable. Referenced as a frozen interface; not modified, criticized, or defended in this memo.

Scope note

This memo treats the international order as the limiting case of a norm-generating system: a system that produces genuine norms (obligations, permissions, prohibitions, powers) but lacks the feature Memo 01 and Memo 07 could take for granted — a centralized enforcement authority holding an effective monopoly on legitimate force. The organizing claim is that international governance is not "weak domestic law" but a structurally distinct regime in which validity, legitimacy, and efficacy dissociate by default rather than by exception, jurisdiction is consent-gated rather than territorial-compulsory, and enforcement is horizontal (self-help, reciprocity, reputation, collective authorization) rather than vertical (a hierarchically superior enforcer). The memo covers: public international law and the sources doctrine of ICJ Statute Art. 38 (treaties, custom = practice + opinio juris, general principles, judicial teachings, and the special category of jus cogens/peremptory norms); international organizations (the UN system, the Security Council veto, the General Assembly); the European Union as a sui generis supranational legal order (direct effect, primacy, the ordinary legislative procedure, the acquis communautaire); the WTO and its dispute-settlement machinery; international human-rights regimes; lex mercatoria and private transnational ordering; and technical standards bodies (ISO, IETF) as de-facto global norm producers. It applies the six-part per-source analysis (origin / legitimacy / enforcement / mutation / hierarchy / conflicts / computational implications) to three sources: international law (as a system), treaties/contracts (as multilateral consent instruments), and transnational private ordering. It closes with a comparison of enforcement models and a Handoff to D1/D2. Political neutrality per §0.7 is observed throughout: the memo describes how these systems produce and enforce norms; it does not rank them or take positions on contested disputes. Deep jurisprudence (Hart, Kelsen, Dworkin) is in Memo 02; the domestic political machine in Memo 01; organizational governance in Memo 07; mutation dynamics and feedback loops in Memo 09; taxonomy in Memo 10; the anti-computability attack (Schmitt, Hayek) in Memo 11.

1. Framing: what changes when the sovereign is removed

Memo 01 modeled a political system as a norm-generating machine NGM = <I, S, A, C, L, E, M, H> (inputs, state, aggregation function, cadence, legitimacy claim, enforcement path, mutation mechanism, hierarchy). This memo reuses that tuple but stresses the component that the domestic memos could treat as reliably present: the enforcement path E.

FACT In the classical account of the modern state (the definition of the state as the holder of a monopoly on the legitimate use of physical force within a territory is the position associated with Max Weber), enforcement is vertical: a hierarchically superior organ (police, courts with compulsory jurisdiction, bailiffs) can apply coercion to a non-compliant subject without that subject's ongoing consent.

FACT The international order has no such organ. There is no world police, no court with general compulsory jurisdiction over states, and no central legislature that can bind a state without some form of its consent. This condition is standardly termed anarchy in international-relations theory — meaning absence of a hierarch, not disorder. INFERENCE "Anarchy" in this technical sense is the single most important structural fact for D1/D2: it means the enforcement subsystem E cannot be assumed, and jurisdiction (the precondition for A and E to apply to a given subject) is itself a variable that must be established per-subject per-norm.

INFERENCE The consequence is a re-ordering of the three observables from Memo 01:

INFERENCE Domestically, validity tends to drag efficacy along (a valid law is usually enforced). Internationally, the link is cut: a norm can be unambiguously valid (a ratified treaty in force) and routinely violated, because the machine that would close the gap — a superior enforcer — does not exist. This is the central contrast with Memo 01 and Memo 07, and it recurs in every section.

1.1 Three questions the domestic model could skip

INFERENCE Removing the sovereign forces three questions to the surface that domestic governance can usually leave implicit:

1. Jurisdiction / applicability. Does this norm apply to this subject at all? Under consent-based jurisdiction the answer is not "yes, because the subject is in the territory" but "only insofar as the subject has consented (by treaty, by acceptance of a court's jurisdiction, by participation in a custom)." Applicability is a computed predicate, not a background assumption.
2. Enforcement authorization. Who, if anyone, may impose a consequence, and under what authority? The answers (self-help, countermeasures, collective authorization, reputational downgrade) are plural and often lack a common apex.
3. Recognition. Who counts as a subject/actor (a "state," an "organization")? Membership in the community of norm-subjects is itself normatively produced (recognition of states, admission to organizations), not given.

INFERENCE Each of these is a place where a compiler built for the domestic case would silently assume a default (applies / a court can enforce / the parties are known) that is false by default internationally. Memo 12 must therefore treat jurisdiction, enforcement-authorization, and recognition as explicit, first-class, defeasible predicates, not ambient facts.

1.2 Vocabulary specific to this memo

• Consent-based jurisdiction — a norm binds a subject only to the extent the subject has consented (by treaty ratification, jurisdictional declaration, or participation in the practice that grounds a custom). Contrast compulsory jurisdiction (domestic default).
• Self-help — a wronged party enforces its own claim by its own action (retorsion, countermeasures, and — at the sharp end — use of force under the UN Charter exceptions), because no superior enforcer will do it for them.
• Reciprocity — compliance sustained by the expectation of matching compliance and the threat of matching defection; the workhorse enforcement mechanism of a horizontal order.
• Reputation — the discounted future cost of being known as a defector; an informational enforcement mechanism (developed in Memo 06 for social norms, here scaled to states and firms).
• Peremptory norm (jus cogens) — a norm from which no derogation is permitted even by agreement; the one place the consent-based order asserts a mandatory, non-optional layer.
• Sui generis — "of its own kind"; used for the EU because it fits neither the "ordinary international organization" nor the "federal state" box.

2. Established consensus: public international law and its sources

Claims here are broadly agreed within public international law and international-relations scholarship. Individual tags qualify each claim; consensus does not upgrade a claim beyond its evidence.

2.1 The sources doctrine (ICJ Statute Art. 38)

FACT Article 38(1) of the Statute of the International Court of Justice enumerates the sources the Court applies, and is treated by most practitioners as the canonical (if not exhaustive) statement of the sources of international law:

Art. 38(1) ICJ Statute — applied by the Court, in this listing:

FACT (a)–(c) are primary sources (they can create obligations); (d) are subsidiary means for determining rules, not independent sources of obligation. INFERENCE For a compiler this is a provenance/type distinction: (a)–(c) are norm-generating operations; (d) is a norm-interpreting/finding operation. They must not share one "source" field.

FACT Art. 38 does not establish a hierarchy among (a), (b), and (c): treaty and custom are, as sources, of equal rank; a later treaty can override an earlier custom between the parties and a later custom can override an earlier treaty. INFERENCE There is therefore no built-in lex superior among the primary sources — conflict is resolved (when it is resolved at all) by lex posterior, lex specialis, and consent, not by a source-rank ordering. This is a sharp contrast with the domestic constitution > statute > regulation hierarchy of Memo 01.

2.2 Treaties (conventions)

FACT A treaty is an international agreement concluded between states (or between states and international organizations) in written form and governed by international law; the codifying instrument for treaties between states is the Vienna Convention on the Law of Treaties (VCLT, concluded 1969, in force 1980). FACT The foundational operative maxim is pacta sunt servanda — agreements in force are binding upon the parties and must be performed in good faith (VCLT Art. 26). INFERENCE A treaty is structurally a multilateral contract: it binds only its parties (the principle pacta tertiis nec nocent nec prosunt — a treaty neither harms nor benefits third parties, VCLT Art. 34), it is entered by an explicit consent act (signature + ratification), and it can carry reservations (a party's unilateral exclusion/modification of certain provisions as to itself, VCLT Arts. 19–23).

INFERENCE The reservation mechanism is computationally striking: it means a multilateral treaty is often not a single shared norm set but a matrix of pairwise-variant norm sets — party A and party B may be bound to each other on terms that differ from A–C. The "same" treaty is a family of bilateral relations. A compiler cannot model a treaty as one global object; it must model a per-pair (or per-party) effective obligation set.

FACT Treaties can terminate or be suspended by their own terms, by consent, by material breach (VCLT Art. 60), by supervening impossibility (Art. 61), or by fundamental change of circumstances (rebus sic stantibus, Art. 62, narrowly construed). INFERENCE Material breach as a ground for the other party to suspend/terminate is a reciprocity primitive written into the contract law itself: defection licenses proportionate defection.

2.3 Customary international law (CIL)

FACT Customary international law arises from two cumulative elements: (i) a general and consistent practice of states (the objective/material element), and (ii) opinio juris sive necessitatis — the belief that the practice is engaged in out of a sense of legal obligation (the subjective element). Both are required; mere habit without opinio juris is not law, and a claimed obligation without practice is not custom. This two-element test is associated with the jurisprudence of the ICJ (e.g., the North Sea Continental Shelf cases) and is codified in substance by the International Law Commission's conclusions on identifying CIL.

INFERENCE CIL is the paradigm of a norm-from-practice generator (Memo 10's "norm generated by consensus/tradition" family) operating at the state level: it is emergent, unwritten, has no enactment event and no promulgation timestamp, and its existence is a matter of evidence and inference rather than of a validity stamp. INFERENCE This is the hardest source for a compiler: the norm has no canonical text, no author, and no effective date; its content and even its existence are contested and probabilistic.

FACT A state that persistently objects to an emerging customary rule during its formation, and does so openly and consistently, is (on the majority view) not bound by that rule once it crystallizes — the persistent objector doctrine. INFERENCE This is a consent residue inside a non-consensual-looking source: even custom, which appears to bind all states, contains an opt-out keyed to timely objection. Computationally it is a per-subject exemption predicate attached to an otherwise general norm.

2.4 General principles, and subsidiary means

FACT "General principles of law recognized by civilized nations" (Art. 38(1)(c)) are principles found across major domestic legal systems (e.g., good faith, res judicata, estoppel, that a breach entails a duty to make reparation) imported to fill gaps where treaty and custom are silent. INFERENCE They function as a gap-filling / default layer — the international order's answer to the non liquet problem (a court's inability to say "there is no law here"). For a compiler this is a fallback rule-set invoked when higher-provenance sources under-determine a case.

FACT Judicial decisions and scholarly teachings (Art. 38(1)(d)) are subsidiary means; international courts formally are not bound by stare decisis (the ICJ's decisions have no binding force except between the parties and in respect of the particular case, ICJ Statute Art. 59). INFERENCE In practice courts follow their own prior reasoning closely, so a de-facto precedential pull exists without a de-jure precedent rule — an efficacy/ legitimacy phenomenon rather than a validity rule (contrast the common-law stare decisis of Memo 02).

2.5 Peremptory norms (jus cogens)

FACT A peremptory norm of general international law (jus cogens) is a norm accepted and recognized by the international community of states as a whole as one from which no derogation is permitted and which can be modified only by a subsequent norm of the same character (VCLT Art. 53). A treaty that conflicts with a peremptory norm is void (Art. 53), and a treaty conflicting with a new peremptory norm becomes void and terminates (Art. 64). FACT The commonly cited (though not closed) examples include the prohibitions of aggression, genocide, slavery, and torture. INFERENCE Jus cogens is the one mandatory, non-derogable layer in an otherwise consent-optional system — it is the international order's lex superior, overriding even the free contracting of states. Computationally it is a small set of hard constraints that invalidate conflicting norms regardless of consent; the exact membership of the set is contested (see §6).

FACT Related but distinct are obligations erga omnes — obligations owed to the international community as a whole, such that all states have a legal interest in their protection (the concept is associated with the ICJ's Barcelona Traction dictum). INFERENCE Erga omnes changes the standing question: normally only an injured party may invoke a breach, but for these obligations any state may have standing. It is a rule about who may enforce, orthogonal to jus cogens, which is a rule about what cannot be derogated from.

2.6 The core structural facts (summary)

• FACT International law is primarily consensual: states are bound because they have consented (by treaty) or acquiesced (by custom), with jus cogens the narrow exception.
• FACT There is no compulsory general adjudication: the ICJ's contentious jurisdiction depends on state consent (by special agreement, by treaty clause, or by an optional-clause declaration under Art. 36(2), which states may make with reservations and may withdraw).
• FACT There is no central enforcer: enforcement is decentralized (self-help, countermeasures, reciprocity, reputation) with one partial exception — the UN Security Council's Chapter VII powers (§3).
• INFERENCE Therefore validity, legitimacy, and efficacy are routinely dissociated, and the system's efficacy rests heavily on the interest structure of the parties rather than on a coercive backstop.

3. International organizations: partial vertical structure inside a horizontal order

FACT International organizations (IOs) are entities established by treaty among states, possessing their own legal personality and organs, exercising powers conferred by their member states. INFERENCE An IO is an attempt to bolt a limited vertical structure onto the horizontal order: members consent, once, to a standing machine that can then produce norms or decisions with some autonomy from case-by-case consent. The degree of that autonomy is the key variable.

3.1 The United Nations: structure

FACT The UN Charter (1945) establishes principal organs including the General Assembly (GA, plenary, all members), the Security Council (SC, 15 members: 5 permanent — China, France, Russia, the UK, the US — and 10 elected), the International Court of Justice (principal judicial organ), the Secretariat, and the (now dormant) Economic and Social and Trusteeship Councils.

3.2 The General Assembly: one-state-one-vote, non-binding output

FACT In the GA each member has one vote (Charter Art. 18), regardless of population, wealth, or power; important questions require a two-thirds majority. FACT With narrow exceptions (the UN's internal matters — budget/apportionment under Art. 17, admission and election of members, organ composition), GA resolutions are recommendations and are not legally binding on member states.

INFERENCE The GA is a sovereign-equality aggregation function (1 state = 1 vote) whose output is, for most purposes, advisory. This decouples validity of the vote from bindingness of the output: the resolution is validly adopted yet imposes no obligation. INFERENCE GA resolutions nonetheless matter as evidence of opinio juris feeding CIL formation (§2.3) and as legitimacy signals — a norm-shaping rather than norm-binding role. A compiler must be able to represent an output that is validly produced but non-binding, and separately track its evidential weight toward custom.

3.3 The Security Council: binding decisions and the veto

FACT Under Charter Art. 25 members agree to accept and carry out SC decisions; under Chapter VII the SC may determine the existence of a threat to the peace and impose binding measures, including sanctions (Art. 41) and authorization of the use of force (Art. 42). FACT SC decisions on non-procedural ("substantive") matters require nine affirmative votes including the concurring votes of the permanent members (Art. 27(3)) — the veto: any one of the P5 can block a substantive decision. (By long practice, an abstention by a permanent member does not count as a veto.)

INFERENCE The SC is the only organ in the general international system with a reliable vertical, binding output — the nearest thing to a world executive with coercive authorization. But its aggregation function A contains a five-way single-point-of-veto gate: the machine can only fire when none of five designated actors objects. INFERENCE Computationally this is a unanimity-among-privileged-members constraint superimposed on a qualified majority — a conjunction of a majority predicate and five individual veto predicates. It makes SC output highly available for blocking and highly scarce for action: the enforcement path exists but is gated behind a five-key lock.

INFERENCE The veto is the clearest illustration of the memo's theme: the international order purchased a binding enforcement organ only at the price of making it structurally paralyzable by any great power. This is a deliberate design trade (great-power buy-in in exchange for great-power vetoes), not a bug — and it means the strongest enforcement channel is conditional on great-power concurrence, i.e., not applicable against a permanent member or its protégés.

3.4 Conferred powers, ultra vires, and implied powers

FACT IOs possess only the powers conferred on them by their constituent treaty (the principle of attributed/conferred competence); acts exceeding those powers are ultra vires. FACT Alongside express powers, IOs are recognized to have implied powers — those necessary to carry out their functions — a doctrine associated with the ICJ's Reparation for Injuries advisory opinion (recognizing the UN's legal personality and capacity to bring claims). INFERENCE Conferred-competence is a capability model: the organization is a machine whose permissible operations are enumerated by its charter, with a (contested) penumbra of implied operations. For a compiler this maps to an explicit competence/permission set per organ, plus a defeasible "necessary-and-implied" extension predicate — and a way to flag an act as ultra vires (validity failure by competence, not by procedure).

3.5 Other IO decision rules (for contrast)

• FACT Many technical/financial IOs use weighted voting keyed to contributions (e.g., the voting structures of the IMF and World Bank allocate votes by quota/shares), rather than one-state-one-vote. INFERENCE This is a different aggregation function — stake-weighted rather than sovereign-equal — and it changes who can form a blocking minority.
• FACT Some bodies operate by consensus (decision without formal vote unless a member objects), including much WTO decision-making (§5) and many standards bodies (§8). INFERENCE Consensus is a near-unanimity aggregation that maximizes buy-in/legitimacy at the cost of decision latency and susceptibility to holdouts.

4. The European Union: a sui generis supranational legal order

FACT The EU is founded on treaties (currently the Treaty on European Union, TEU, and the Treaty on the Functioning of the European Union, TFEU) concluded among sovereign member states, yet it exhibits features that ordinary international organizations lack. The Court of Justice of the European Union (CJEU) characterized the (then) Community as "a new legal order of international law" for whose benefit the states limited their sovereign rights (the position established in Van Gend en Loos, 1963), and asserted that this order forms an integral part of the member states' own legal systems. INFERENCE The EU is best modeled as a third category: neither a treaty-bound IO whose norms bind only states and require domestic transposition, nor a federal state with its own coercive apparatus, but a hybrid that penetrates directly into member legal orders while still resting on a treaty base that members could (per Art. 50 TEU) exit.

4.1 Direct effect

FACT Direct effect is the doctrine (originating in Van Gend en Loos) that provisions of EU law which are sufficiently clear, precise, and unconditional can be invoked by individuals directly before national courts, without the need for national implementing measures. FACT Regulations are directly applicable by their nature (TFEU Art. 288); Treaty provisions and directives can have direct effect under conditions — with directives recognized to have vertical direct effect (individual against the state) but, on the CJEU's case law, generally not horizontal direct effect (individual against individual).

INFERENCE Direct effect changes the subject set of the norm. Classical international law binds states, which then may (or may not) create domestic rights; EU law can create rights and duties for individuals and firms directly, enforceable in the member's own courts. Computationally the norm's subject field expands from {states} to {states, natural persons, legal persons}, and enforcement is routed through the member states' own vertical enforcement machinery — the EU borrows the members' monopolies on force rather than building its own. This is the mechanism by which a treaty-based order acquires domestic-grade efficacy.

4.2 Primacy (supremacy)

FACT Primacy is the doctrine (established in Costa v ENEL, 1964) that EU law takes precedence over conflicting national law, including — on the CJEU's view — conflicting national constitutional provisions (Internationale Handelsgesellschaft). FACT This claim is contested at the boundary by several national constitutional courts, which assert a reserved competence to review EU acts for compatibility with core constitutional identity or for acting ultra vires (the line of reasoning associated with the German Federal Constitutional Court's Solange, Maastricht, Lisbon, and PSPP/Weiss jurisprudence).

INFERENCE Primacy supplies a scoped lex superior: within the field of conferred EU competence, EU norms outrank national norms. But the ordering is not a single accepted total order — the CJEU's account (EU law supreme, full stop) and the national-constitutional-court account (supreme unless it breaches constitutional identity / is ultra vires) disagree about the apex. INFERENCE For a compiler this is the canonical case of two overlapping hierarchies with contested tops: a scoped supremacy predicate plus a defeasibility clause (national-identity / competence review) whose activation is itself disputed. It must be representable as an unresolved conflict of hierarchies, not as a fabricated single ordering (cf. Memo 01 §6.6 "no shared apex").

4.3 The ordinary legislative procedure (former co-decision)

FACT The ordinary legislative procedure (OLP; before the Lisbon Treaty, "co-decision") is the EU's principal law-making process (TFEU Art. 294): the Commission proposes; the European Parliament (directly elected) and the Council (member-state ministers) act as co-legislators who must both agree on the same text, across up to three readings, with a Conciliation Committee if they diverge. The Council in most OLP matters votes by qualified majority voting (QMV).

ORDINARY LEGISLATIVE PROCEDURE (simplified control flow)

FACT QMV under the current "double majority" rule requires (broadly) 55% of member states representing at least 65% of the EU population (TEU Art. 16(4)). INFERENCE The OLP is a bicameral aggregation combining a directly-elected chamber (population-legitimacy) with a state chamber (sovereign-legitimacy), and its QMV rule is a dual-threshold function (count of states AND share of population) — a deliberately two-dimensional majority intended to prevent either large or small states from dominating. Computationally this is a conjunction of two weighted-majority predicates over different weightings of the same voters — considerably richer than the SC's veto or the GA's simple one-state-one-vote.

4.4 The acquis communautaire

FACT The acquis communautaire ("the Community acquis") is the accumulated body of EU law — treaties, legislation, judicial decisions, and principles — that binds all members and that acceding states must adopt in full as a condition of membership. INFERENCE The acquis is a monotonically accumulating, non-negotiable-at-entry state S: a new member does not renegotiate the existing norm set; it imports the entire store as a precondition. Computationally this is an "accept-the-whole-repository-to-join" model — the opposite of the per-treaty, reservation-friendly, pick-your-obligations model of classical international law (§2.2). It is also a strong path-dependence / ratchet feature relevant to Memo 09's mutation dynamics: the acquis grows and is very hard to shrink.

4.5 Why the EU is the interesting middle case

INFERENCE The EU matters to this series because it sits between the domestic machine of Memo 01 (vertical enforcement, clear apex) and the horizontal order of §2–§3 (consent, self-help, no apex). It shows that the sovereign/no-sovereign distinction is a spectrum, and that a treaty order can approach domestic-grade efficacy by (i) creating individual-level subjects (direct effect), (ii) asserting scoped supremacy (primacy), and (iii) borrowing the members' enforcement monopolies rather than building a central one — while remaining vulnerable at exactly the point where the borrowed hierarchy's apex is contested (§4.2). HYPOTHESIS The EU is thus the empirical demonstration that "amount of sovereign" is a tunable parameter, not a binary — a claim D1/D2 should be able to represent as a degree of verticality rather than a flag.

5. The WTO and dispute settlement: rules with teeth, then without

FACT The World Trade Organization (established 1995, succeeding the GATT 1947 regime) administers a package of trade agreements binding on its members, and operates a Dispute Settlement Understanding (DSU). Under the DSU, a complaining member may request consultations and then a panel; panel reports could be appealed to a standing Appellate Body; and adopted rulings, if not complied with, can lead to authorized retaliation — the suspension of trade concessions by the winning member against the losing member, up to a level determined by arbitration.

FACT A defining innovation of the DSU was the negative (reverse) consensus rule for adopting panel/Appellate Body reports: a report is adopted unless there is consensus against adoption. INFERENCE This inverts the veto logic of the Security Council. Under positive consensus, any interested party (including the loser) can block; under negative consensus, the loser would have to persuade everyone, including the winner, to block — which is practically impossible. Computationally, negative consensus converts a blockable step into a near-automatic one: the default flips from "no unless all agree" to "yes unless all object." This gave WTO adjudication an efficacy closer to a domestic court than almost anything else in international law.

FACT The Appellate Body ceased to function in December 2019 when the United States blocked the appointment/reappointment of its members, leaving it without the quorum needed to hear appeals; members have since used interim arrangements (e.g., the Multi-Party Interim Appeal Arbitration Arrangement) and "appeals into the void." INFERENCE This is a live demonstration of the memo's core fragility: even the strongest international enforcement machine depended on an appointments process gated by member consent, and a single powerful member's refusal to consent disabled the apparatus without amending a single rule. The norms remained valid; the enforcement path lost efficacy. It is the international analogue of a court that still exists on paper but can no longer seat judges.

INFERENCE The WTO enforcement model — authorized proportionate retaliation — is structured self-help: the system does not enforce its own judgments; it licenses the winner to enforce by withdrawing equivalent concessions, with the level policed by arbitration. This is reciprocity formalized and capped. INFERENCE It works best among parties with roughly symmetric trade stakes (mutual vulnerability makes the threat credible) and works poorly where stakes are asymmetric (a small economy's authorized retaliation against a large one is a weak threat). Enforcement efficacy is therefore a function of the parties' relative exposure, not a constant — another variable a compiler cannot treat as fixed.

6. International human-rights regimes

FACT International human-rights law is embodied in treaties (e.g., the two 1966 Covenants on civil-political and on economic-social-cultural rights; specialized conventions on torture, racial discrimination, the rights of the child, etc.) and in regional systems with courts — notably the European Court of Human Rights (ECtHR) under the European Convention on Human Rights, and the Inter-American and African regional systems. FACT Enforcement ranges from reporting/monitoring by treaty bodies (which issue non-binding "views"/recommendations) to binding regional-court judgments (the ECtHR's judgments are binding on respondent states, with supervision of execution by the Council of Europe's Committee of Ministers).

INFERENCE Human-rights regimes are the sharpest example of the validity/efficacy gap: the norms are among the most widely ratified in existence (high validity), yet compliance is uneven and enforcement often reduces to reporting, naming, and reputational pressure rather than coercion. The regional courts are the exception that proves the rule — they achieve real efficacy precisely where states have consented to a standing court plus a political supervision organ, i.e., where a partial vertical structure was built.

INFERENCE Human-rights norms also stress the subject/standing question: they are largely obligations of a state toward individuals within its jurisdiction, and some overlap with jus cogens / erga omnes (§2.5), meaning breaches may be invocable by actors other than a directly injured state. OPEN Whether "human rights" as content can be represented in a content-neutral kernel is exactly the kind of value-laden question JD Q8 flags for exclusion above the kernel; this memo treats human-rights content as above-kernel and only its structural features (subject expansion, weak enforcement, reporting mechanisms, regional courts) as candidate interface features. (See §9 and Memo 12.)

FACT Human-rights treaties frequently contain derogation clauses (permitting suspension of some rights in a declared public emergency, subject to conditions) while designating certain rights as non-derogable (e.g., the prohibition of torture). INFERENCE This is a built-in state-conditioned modality switch: the applicable norm set changes as a function of a declared state (emergency), with a protected core that the switch cannot touch — structurally identical to the jus cogens "hard constraint" pattern (§2.5) and worth modeling as a general "emergency mode with a non-suspendable kernel of constraints."

7. Lex mercatoria and private transnational ordering

FACT Lex mercatoria ("law merchant") refers to a body of transnational commercial norms — trade usages, standard-form contracts, model clauses, and principles (e.g., the UNIDROIT Principles of International Commercial Contracts; INCOTERMS published by the International Chamber of Commerce) — generated and applied largely by the commercial community itself, often chosen by parties to govern their contracts and applied in international commercial arbitration. FACT Its status as law is contested: some regard it as genuine autonomous law, others as a set of usages that bind only via the parties' contractual choice and the enforcing state's recognition.

FACT The enforcement backbone of this private order is arbitration plus a treaty for recognizing awards: the New York Convention (Convention on the Recognition and Enforcement of Foreign Arbitral Awards, 1958) obliges its many state parties to recognize and enforce foreign arbitral awards, subject to narrow exceptions (e.g., public policy). INFERENCE This is the crucial hybrid: a privately produced norm and privately adjudicated dispute is ultimately made efficacious by borrowing the states' domestic enforcement monopolies through a recognition treaty — exactly the EU's "borrow the members' coercion" pattern (§4.1), but organized around private commercial actors instead of an integration project.

INFERENCE Private transnational ordering demonstrates that a norm system can achieve high efficacy with no public law-maker at all, provided it can (i) generate norms by community practice and contract, (ii) adjudicate by consent (arbitration), and (iii) plug into an enforcement layer it does not own. Reputation is a large part of the pre-award enforcement: merchants, industries, and diamond/commodity trades sustain compliance through repeat-play reputational sanctions and exclusion from the network (a mechanism analyzed at the social level in Memo 06 and the institutional level in Memo 05). HYPOTHESIS The general pattern — private generation + consensual adjudication + rented public enforcement + reputational backstop — is a reusable architecture that may recur wherever a public sovereign is absent or declined (including, plausibly, digital/on-chain ordering), and is worth treating as a first-class "enforcement topology" in D1/D2.

8. Technical standards bodies: de-facto global norms

FACT Standards-development organizations — e.g., ISO (International Organization for Standardization), the IEC, the ITU, and internet-governance bodies such as the IETF (which produces RFCs), the W3C, and ICANN — produce technical standards that are formally voluntary yet frequently become de-facto mandatory through market adoption, network effects, procurement requirements, or incorporation-by-reference into binding law and regulation. FACT The IETF's operating ethos is captured in the (frequently quoted) formula associated with David Clark — "rough consensus and running code" — and its norms (RFCs) are adopted through open participation and working-group consensus rather than state votes.

INFERENCE Standards bodies are a norm generator of the "consensus + demonstrated implementation" family (Memo 10): a norm becomes authoritative not by enactment or by state consent but by (i) working-group rough consensus and (ii) evidence that it works (running code / interoperable implementations). This couples a consensus aggregation to an empirical validity test — a norm is "real" partly because it demonstrably functions. That empirical gate has no analogue in the treaty/custom sources and is of direct interest to a program called Computable Governance.

INFERENCE Their enforcement is almost purely market + interoperability + reputation: there is no sanction for ignoring a standard except failing to interoperate and being excluded from the network. Yet efficacy can exceed that of ratified treaties, because the cost of non-compliance (incompatibility) is immediate, automatic, and self-imposed rather than requiring an enforcer. HYPOTHESIS "Norm enforced by loss of interoperability" is a distinct enforcement primitive — automatic, decentralized, and requiring no adjudicator — and may be the cleanest naturally-computational enforcement model in the whole survey. INFERENCE Standards also show de-facto normativity without formal validity: an RFC is not "law," has no sovereign, yet governs global behavior — a direct challenge to any validity-centric model and a strong data point for the JD Q7 relations among Technology / Standards / Power / Trust.

FACT Standards frequently acquire hard legal force by incorporation by reference: a statute or regulation makes compliance with an external standard legally mandatory (e.g., safety codes referencing ISO/IEC standards). INFERENCE This is a cross-source promotion event: a privately-produced voluntary norm is lifted into the binding legal hierarchy by a public instrument pointing at it. Computationally it is a reference/linking operation between norm stores of different provenance — the compiler must support a norm whose content lives in one system and whose bindingness is conferred by another.

9. Six-part per-source analysis (JD Q2)

Per §0.5 and JD Q2, each source is analyzed on origin / legitimacy / enforcement / mutation / hierarchy / conflicts / computational implications. Three sources are treated: international law (as a system), treaties/contracts, and transnational private ordering.

9.1 Source: International law (as a system)

• Origin. FACT Emerges from the horizontal interaction of formally equal sovereigns via the Art. 38 sources — express agreement (treaties), practice + opinio juris (custom), imported common principles, with a narrow non-derogable core (jus cogens). No constituent moment, no founding sovereign; the system bootstraps from the mutual recognition of states.
• Legitimacy. INFERENCE Grounded in consent and reciprocity (states are bound because they agreed or acquiesced) supplemented by fairness/community claims for the jus cogens/erga omnes core. Legitimacy is plural and partial — no single accepted ground — and is sharply decoupled from validity and efficacy.
• Enforcement mechanism. FACT Decentralized: self-help (retorsion, countermeasures), reciprocity (material-breach responses), reputation, collective authorization (SC Chapter VII, when the veto permits), and third-party adjudication only on consent. No monopoly on force; the one binding coercive channel (SC) is veto-gated.
• Mutation mechanism. FACT Treaties change by amendment/consent; custom changes by evolving practice + opinio juris (slow, unwritten, contested onset — see Memo 09); jus cogens changes only by a subsequent norm of the same character. INFERENCE Mutation is multi-track and mostly non-legislative — there is no central editor of the corpus.
• Hierarchy. FACT Flat among primary sources (treaty ≈ custom ≈ principles), with a single superior layer (jus cogens) and a standing-expander (erga omnes). INFERENCE There is no general apex validity predicate; ordering is by lex posterior/specialis + consent.
• Conflicts. FACT Resolved (when at all) by lex specialis, lex posterior, treaty interpretation (VCLT Arts. 31–32), and self-contained regimes; many conflicts simply remain unresolved (no court with jurisdiction). INFERENCE UNRESOLVED is a normal terminal state, not an error.
• Computational implications. INFERENCE Jurisdiction/applicability must be a per-subject per-norm computed predicate; validity/legitimacy/efficacy must be three independent fields with efficacy routinely below validity; enforcement is an optional, attachable, plural subsystem; the conflict resolver must return UNRESOLVED honestly; and jus cogens is a small set of hard invalidating constraints of contested membership.

9.2 Source: Treaties / contracts (multilateral consent instruments)

• Origin. FACT A deliberate consent act by the parties: negotiation, adoption, signature, and ratification (bringing the treaty into force per its own terms), governed by the VCLT. Origin is event-stamped and authored — unlike custom.
• Legitimacy. INFERENCE Pacta sunt servanda + explicit consent: the party is bound because it chose to be. Legitimacy is strongest here of any international source because it reduces to self-binding. Reservations mean consent is granular (per-provision).
• Enforcement mechanism. FACT Reciprocity written into the instrument (material breach → suspension/termination, VCLT Art. 60); dispute clauses may confer adjudication or arbitration by consent; otherwise self-help and reputation. INFERENCE The contract carries its own (mostly reciprocal) enforcement logic; external coercion is not assumed.
• Mutation mechanism. FACT Amendment by the parties' agreement; modification inter se; termination/withdrawal per the treaty's clauses or the VCLT grounds (breach, impossibility, rebus sic stantibus). INFERENCE Mutation is contractual and party-controlled.
• Hierarchy. FACT A treaty cannot validly derogate from jus cogens (void, VCLT Art. 53); among treaties, later/more-specific generally prevails between the same parties; UN Charter obligations prevail over other treaty obligations for UN members (Charter Art. 103). INFERENCE So there are a few superior anchors (jus cogens; Art. 103), but otherwise the order is party- and time-relative, not a fixed rank.
• Conflicts. FACT Between successive treaties on the same subject: VCLT Art. 30 rules plus lex posterior/specialis; between a treaty and third parties: pacta tertiis (no effect without consent); reservations produce a matrix of variant bilateral obligations. INFERENCE Conflict resolution is largely pairwise, not global.
• Computational implications. INFERENCE Model a treaty as a per-party / per-pair effective obligation set (because of reservations), not one global object; carry an explicit in-force predicate keyed to ratification and the treaty's own entry conditions; represent withdrawal/termination as first-class state transitions; encode Art. 103 and jus cogens as overriding anchors; treat third-party non-effect as a default that consent can lift.

9.3 Source: Transnational private ordering (lex mercatoria, arbitration, standards)

• Origin. FACT Generated by the relevant community: merchants (usages, model clauses), standards working groups (RFCs, ISO standards), and party contracts choosing the governing rules. No public law-maker; origin is practice + private drafting + choice-of-law.
• Legitimacy. INFERENCE Grounded in party autonomy (they chose it), expertise (technical merit / "running code"), and community acceptance. Legitimacy is functional/reputational rather than sovereign.
• Enforcement mechanism. FACT Arbitration + New York Convention recognition (rented state coercion); reputation and exclusion from the network; for standards, loss of interoperability. INFERENCE Enforcement is decentralized and often automatic (incompatibility is self-executing), with public coercion available only at the recognition stage.
• Mutation mechanism. FACT New editions/versions (ISO revision cycles, new RFCs superseding old ones), evolving usages, updated model contracts. INFERENCE Mutation is versioned and often rapid — closer to software release cadence than to treaty amendment.
• Hierarchy. INFERENCE Ordering is mostly by party choice and by version (later edition supersedes earlier); public ordre public / mandatory law sits above it (an award contrary to public policy may be refused enforcement). Standards can be promoted into the legal hierarchy by incorporation-by-reference (§8).
• Conflicts. INFERENCE Between competing standards: resolved by market adoption / network effects, not adjudication; between private norm and mandatory public law: public law wins at the enforcement gate. FACT The public-policy exception is the standing override.
• Computational implications. INFERENCE Support versioned norm objects with supersession links; model "enforced by loss of interoperability" and "enforced by reputational exclusion" as distinct enforcement primitives; support cross-source promotion (incorporation-by-reference) as a linking operation; represent the public-policy/mandatory-law override as a ceiling on private ordering's validity.

10. Comparison of enforcement models

The distinguishing feature of international governance is its enforcement topology. The following table compares the enforcement models encountered in this memo (and, for contrast, the domestic model of Memo 01/07). "Vertical" = a hierarchically superior enforcer can coerce without ongoing consent; "Horizontal" = enforcement is by peers/self.

INFERENCE Reading the table top to bottom is reading a gradient from vertical to horizontal enforcement, and the striking result is that efficacy does not track verticality monotonically: purely horizontal, no-apex models (technical standards; strong arbitration regimes) can achieve efficacy exceeding that of formally binding, widely-ratified treaties. INFERENCE The explanatory variable is not "is there a sovereign enforcer" but the cost and automaticity of the consequence to the defector: interoperability loss is immediate and self-executing; treaty breach's consequence is slow, discretionary, and often absorbable. This is the memo's single most important finding for D1/D2.

10.1 Sharp contrast with Memo 01 and Memo 07

• INFERENCE Memo 01/07 assume E (enforcement) is present and vertical. This memo shows E is the variable: optional, plural, often horizontal, sometimes borrowed, sometimes absent. A compiler tuned to the domestic case would default-assume enforcement and be wrong by default here.
• INFERENCE Memo 01/07 assume jurisdiction from territory/membership. Here jurisdiction is consent-gated and per-norm — an explicit predicate that can be false even for an in-force norm and even for a subject "inside" the system.
• INFERENCE Memo 01/07 have a clear apex validity predicate (rule of recognition / constitution). Here there is no general apex: sources are flat, hierarchies are scoped and contested (EU primacy), and UNRESOLVED is a normal output.
• INFERENCE Memo 01/07 let validity drag efficacy along. Here the three observables are independent by default, and the normal condition is valid but under-enforced.
• INFERENCE Consequence: the domestic memos describe a machine; this memo describes a network of machines with no shared operating system, coordinating by consent, reciprocity, and reputation. Memo 12's interface must be expressive enough to cover both — which means the domestic case is a special (fully-vertical, single-apex, compulsory-jurisdiction) configuration of the more general international case, not the other way around.

10.2 Normative feedback loops (JD Q5) specific to the international layer

INFERENCE Loops that D1/D2 and Memo 09 should track:

INFERENCE Loops (b), (d), and (e) are all cross-source promotion loops: a norm migrates from a weaker/optional provenance to a stronger/binding one without being re-enacted. This is distinctive to the international layer (domestic systems mostly promote by explicit enactment) and is the strongest argument that provenance must be a mutable attribute over time, not a fixed birth-stamp — a point handed to Memo 09.

11. Competing theories

Per §0.5, live disagreements are presented without adjudication.

• Is international law "really law"? FACT A classic challenge (the position associated with John Austin's command-theory jurisprudence) held that international law is not law "properly so called" but "positive morality," because it lacks a sovereign issuing enforceable commands. FACT The rival position (associated with H. L. A. Hart) accepts international law as law while noting it may lack a developed rule of recognition and secondary rules — i.e., it is law with an underdeveloped validity apparatus, not non-law. OPEN The dispute is unresolved and matters directly to the kernel: if "law requires a sovereign enforcer," the international layer falls outside a validity-centric model; if not, it is law with a different enforcement topology. This memo takes no side and models both the norms and the enforcement gap.
• Consent vs. objective sources. FACT Voluntarist/positivist accounts ground all international obligation in state consent; naturalist / community accounts ground at least jus cogens and erga omnes in values independent of consent. OPEN Whether the non-derogable core is "discovered" (objective) or "agreed by the community as a whole" (a super-consent) is contested — and determines whether the hard-constraint set is fixed or itself consent-mutable.
• Realism vs. institutionalism (compliance). FACT Realist international-relations theory treats international norms as epiphenomenal to power and interest (states comply when it suits them); institutionalist / regime theory holds that institutions independently raise compliance by lowering transaction costs, providing information, and enabling reciprocity. OPEN The causal weight of norms vs. interest in producing compliance is empirically contested; it bears on whether "efficacy" is a property of the norm/enforcement system or merely of the underlying interest alignment.
• Monism vs. dualism (international/domestic relation). FACT Monism treats international and domestic law as one system in which international law is directly effective; dualism treats them as separate, requiring domestic transposition for international law to have internal effect. INFERENCE States sit at different points on this axis, and the EU's direct effect is a strong-monist design choice; a compiler cannot assume either — the international/domestic linkage must be a configurable coupling, not a constant.
• Fragmentation vs. constitutionalization. FACT One view stresses fragmentation — the proliferation of specialized, self-contained regimes (trade, human rights, environment, law of the sea) with their own tribunals and no apex, risking conflicting rulings. Another stresses an emerging constitutionalization of international law around jus cogens, the Charter, and shared principles. OPEN Which trend dominates is unsettled and directly affects whether a single coherent norm graph is even the right target for D1.

12. Open questions

• OPEN No apex validity predicate. Without a rule of recognition, how does a compiler decide whether a putative customary norm exists? Its existence is evidential and contested; there may be no fact of the matter at a given moment, only a probability.
• OPEN Membership of jus cogens. The non-derogable set is agreed to exist but its exact contents are contested and possibly growing. A hard-constraint layer of uncertain, mutable membership is hard to formalize without begging the question.
• OPEN Measuring efficacy, legitimacy, reputation. Compliance, legitimacy, and reputational cost are noisy, inferred, and probably not cardinal — yet the whole horizontal-enforcement story depends on them. No agreed metrics (shared with Memo 01 §9).
• OPEN Onset of custom. Custom has no enactment timestamp; the moment a practice "crystallizes" into law is genuinely indeterminate, as is the closing of the persistent-objector window. Representing a norm with a fuzzy birth date is unsolved.
• OPEN Contested overlapping hierarchies. EU primacy vs. national constitutional identity (§4.2) has no agreed apex; the same shape recurs across fragmented regimes. Whether a compiler should ever synthesize a resolution, or must always return UNRESOLVED, is a design decision with honesty (M1/M6) implications.
• OPEN Where does the private/public boundary fall? Standards and lex mercatoria generate effective global norms with no sovereign; whether these are "governance" for kernel purposes, or above-kernel content, is unsettled and bears on JD Q8 exclusions.

13. Research opportunities (for D1/D2)

Per §0.5, where the compiler work could contribute or must decide.

• INFERENCE First-class jurisdiction/applicability predicate. Make "does norm N apply to subject X?" a computed, defeasible predicate (consent, membership, participation, persistent objection) rather than an ambient truth. This generalizes the domestic case (where it is trivially true) rather than special-casing it.
• INFERENCE Enforcement topology as a typed, attachable field. Enumerate topologies (vertical | veto_gated_vertical | borrowed_vertical | structured_self_help | reciprocity | reputation | interoperability_loss | none) and attach one (or several) to each norm. The domestic monopoly-on-force case is one value among many.
• INFERENCE UNRESOLVED as a first-class conflict result. Required for no-apex conflicts and contested overlapping hierarchies; forbidding it would force fabrication (violates M6).
• INFERENCE Provenance as mutable over time. Support cross-source promotion (soft-law → custom; standard → incorporated law; treaty → custom) as explicit transitions on a norm's provenance, not immutable birth-stamps (§10.2; handoff to Memo 09).
• INFERENCE Per-pair/per-party obligation matrices. Model reservation-bearing multilateral treaties as families of bilateral obligation sets, not single global objects.
• HYPOTHESIS A "degree of verticality" scalar. The EU shows sovereign-ness is a spectrum; representing verticality as a degree (rather than a boolean) may unify domestic and international configurations under one model.
• HYPOTHESIS Interoperability-loss as the canonical computational enforcement primitive. Because it is automatic, decentralized, and adjudicator-free, it may be the enforcement model that maps most cleanly onto executable governance — worth prototyping first in D2.

14. Handoff to D1/D2

Concrete implications for the compiler architects (all INFERENCE unless noted):

1. Do not assume enforcement. The enforcement subsystem E must be an optional, typed, attachable binding (see topologies in §13). Default = none, not present. The domestic monopoly-on-force case (Memo 01/07) is the special case; the international horizontal case is the general one.
2. Jurisdiction is a predicate, not a background fact. Every norm-to-subject application must pass an explicit, defeasible applies(N, X) check (consent, membership, participation, minus persistent objection). It may be false for an in-force norm.
3. Three independent status fields, with efficacy routinely below validity. Carry {validity, legitimacy, efficacy} separately (as in Memo 01) and expect their dissociation to be the normal case internationally, not an exception.
4. No forced apex. Support flat sources (treaty ≈ custom ≈ principles), a small set of overriding anchors (jus cogens, UN Charter Art. 103), and scoped, possibly-contested supremacy (EU primacy with a national-identity defeasibility clause). The conflict resolver must return UNRESOLVED honestly.
5. Consent granularity. Support per-provision consent (treaty reservations) yielding per-pair effective obligation sets; support opt-out keyed to timing (persistent objector) as a per-subject exemption predicate.
6. Cross-source promotion as an operation. Provide transitions that change a norm's provenance over time (soft-law → custom, standard → incorporated-by-reference law, treaty → custom) without re-enactment; provenance is mutable, not a birth-stamp.
7. Model non-binding-but-valid output. GA-style recommendations are validly produced yet impose no obligation while carrying evidential weight toward custom; represent binding = false alongside evidential_weight > 0.
8. Enforcement efficacy is a function of the parties, not a constant. Reciprocity/retaliation models (WTO, treaties) depend on mutual exposure and repeat play; represent the credibility of a consequence as parameter-dependent, not fixed.
9. Borrowed and automatic enforcement are first-class. Support "enforcement borrowed from another system's monopoly" (EU direct effect; New York Convention recognition) and "enforcement by automatic loss of interoperability" (standards) as distinct primitives — the latter may be the cleanest to render executable in D2.
10. Honesty constraints (M1/M6). The compiler must represent: an existent-but-contested norm (custom with no crystallization timestamp), a valid-but-unenforced norm, a contested-apex hierarchy conflict, and a hard-constraint set of uncertain membership (jus cogens) — surfacing each rather than fabricating resolution.
11. Above-kernel content stays out (JD Q8). The content of human-rights norms, the ranking of political systems, and the substance of contested disputes are above the kernel. Only the structural features surveyed here (enforcement topology, jurisdiction predicate, provenance, promotion, overriding anchors, non-binding output, degree of verticality) are candidate interface elements for Memo 12.

FACT , per §0.2 D0 is treated as a frozen interface throughout; nothing above prescribes changes to the kernel. These are inputs the layer above the kernel must supply and the interface (Memo 12) must be able to receive. The decisive lesson of the international case is that the kernel's model of enforcement and jurisdiction must be general enough to include their absence — the domestic sovereign is a configuration, not a premise.

End of Memo 08. Cross-references: Memo 01 (the domestic norm-generating machine and the vertical-enforcement assumption this memo relaxes; the EU and IOs introduced there and deepened here), Memo 02 (Hart/Austin/Kelsen on whether international law is "law" and on the rule of recognition), Memo 05 (contract, reputation, and repeat-play enforcement in institutional economics), Memo 06 (social norms: reputation and exclusion as enforcement), Memo 07 (organizational governance and borrowed/delegated enforcement), Memo 09 (custom formation, cross-source promotion, and feedback-loop dynamics), Memo 10 (norm generators: consent, custom, consensus, running-code), Memo 11 (Schmitt/Hayek/Foucault attacks on computability, and the "exception" that no rule can pre-decide), Memo 12 (the invariant content-neutral interface).