Coupling: Core Conclusions

The coupling series supplies what the coupling literature has never had: an ontological account of what coupling is across all its manifestations. Five decades of dimensions, metrics, and measurement frameworks produced no such account. This page collects what the coupling series concludes.

1 Coupling is access to a shared resource

The central definitional result: two elements are coupled exactly when they access a shared resource.

A resource is anything whose definition, state, or capacity constrains the elements that share it — a module’s interface, a type definition, a database schema, a release pipeline, a team boundary, a deployment unit, a CPU core. A coupling is the ternary relation

(e1,r,e2) (e_1, r, e_2)

— two entities together with the resource that links them. The resource is the medium of the coupling.

The definition is change-independent: coupling is a standing constraint that holds whether or not any change occurs. Interference and shared fate are couplings that need no change at all.

2 The recognized dimensions differ only in the resource

Every recognized coupling dimension — structural, temporal, concurrency, semantic, platform — differs only in which resource is shared and how the sharing is enforced. There is no structural difference underneath.

The series provides a taxonomy of resource types organized by seven constraint-kinds (contract, store, medium, budget, authority, location, time) and three coupling modes (definitional, state, capacity). Each coupling dimension maps to its resource, its governing contract, and its enforcement mechanism.

3 Necessary coupling is to be made manifest, not minimized

The visibility principle: coupling is not to be minimized but made manifest. Each shared resource is governed by an explicit contract whose violation is immediately detectable.

This reframes the classical imperative. The goal is not to eliminate coupling — coupling is a standing constraint of any system that shares resources — but to make it visible and governed. The directive applies to necessary coupling: the shared-resource edges the system genuinely requires. Accidental coupling is a different target, and it is removed, not merely made visible. The two directives are not in tension because they operate on disjoint edge sets — the visibility principle governs the edges the system needs, while the grouping result below removes the edges above the floor.

4 Grouping by identical resource-access minimizes accidental coupling

From the coupling definition alone — without change drivers, Γ, or IVP — one result follows: grouping elements by identical resource-access sets

R(e1)=R(e2) R(e_1) = R(e_2)

minimizes accidental coupling cost.

The series reveals a tripartite structure: E→ℛ→CE \to \mathcal{R} \to C. Every element accesses a set of resources; every resource is governed by a set of change drivers; the element’s change-driver set is derived, not primitive:

Γ(e)=⋃r∈R(e)D(r) \Gamma(e) = \bigcup_{r \in R(e)} D(r)

This is why the coupling-side and cohesion-side derivations are two ends of the same refinement chain: the resource-set partition refines the Γ partition. Γ is derived from RR through the governance mapping.

5 Coupling is a consequence of cohesion

The classical imperative — “maximize cohesion, minimize coupling” — treats the two as independent objectives requiring a trade-off. The series proves otherwise: within any single dimension, coupling is a consequence of cohesion, and the coupling categories that modularization controls are diagnostic of cohesion defects, not independent design variables.

Coupling decomposes into eight mutually exclusive categories — the product of three binary partitions (internal/external, same-class/cross-class, necessary/not), parameterized by a partitioning rule. At maximal cohesion (purity = completeness = 1), all modularization-controlled coupling categories are empty.

The result: the normative decomposition simultaneously maximizes cohesion and minimizes avoidable coupling — one optimum, not two.

6 The coupling floor

There is an irreducible coupling the domain demands — the coupling floor. At correct boundaries, the floor equals the necessary dependency graph exactly; every edge above it is a boundary defect or an accidental edge.

Because coupling is measured against a partitioning rule, necessary coupling is rule-relative: the same edge that one concern demands be internal is, under another, demanded external. The imperative to “minimize coupling” has no fixed referent until the rule is fixed. A raw count cannot separate domain-mandated coupling from placement error.

7 Where the proofs live

The proofs — that coupling is access to a shared resource, that the resource-set partition uniquely minimizes accidental coupling, that coupling is a consequence of cohesion, the coupling floor — are in the published coupling papers: the definition of coupling as shared resource access in the paper on the nature of coupling [1], the derivation of the modularisation directive from coupling in the derivation paper [2], and the coupling floor and coupling-as-consequence-of-cohesion results in the decomposition paper [3].

The formal development. The definitions, theorems, and proofs behind this page are published in:

All published papers

References

[1]
Y. Loth, “On the nature of coupling: All coupling as shared resource access.” Zenodo, Aug. 2026. doi: 10.5281/zenodo.21925122.
[2]
Y. Loth, “A self-contained derivation of the optimal modularization from coupling.” Zenodo, Aug. 2026. doi: 10.5281/zenodo.21925268.
[3]
Y. Loth, “Coupling as a consequence of cohesion: Decomposition, the coupling floor, and domain-imposed coupling.” Zenodo, Aug. 2026. doi: 10.5281/zenodo.21925201.