Standard ML

Standard ML (SML) is a general-purpose functional programming language with an expressive type system, first defined in the mid-1980s by Robin Milner and others as the successor to ML (Meta Language). It is distinguished by its combination of type inference, pattern matching, and higher-order functions, all governed by a strict static type discipline that eliminates entire classes of runtime errors without requiring explicit type annotations. SML is not merely a programming language; it is a formal system in which programs are also mathematical objects amenable to reasoning.

The language's module system — structures, signatures, and functors — is among the most sophisticated of any production language. It permits the specification of abstract interfaces and their generic composition, enabling the construction of large programs from reusable, independently verifiable components. This modularity is not an afterthought; it is foundational to SML's design philosophy, which treats software as a hierarchical structure of theories rather than a collection of instructions.

SML serves as the implementation language for CPN Tools, where its type system is used to define the color sets of Colored Petri Nets and its functional primitives express arc inscriptions and guard conditions. This choice is revealing: the authors of CPN Tools selected a language whose compile-time guarantees mirror the formal guarantees they sought for their models. The pairing of a functional type system with a graphical formalism is a statement about methodology — that visual intuition and formal rigor are not opposites but complements.

Despite its elegant design, Standard ML has never achieved widespread adoption in industry. Its successors — OCaml, Haskell, and the type-theoretic features of modern Rust and TypeScript — have borrowed its innovations while discarding its syntax and ecosystem. SML's academic origins, its relatively small standard library, and the fragmentation of its implementation ecosystem (SML/NJ, MLton, Poly/ML) have confined it to research and education. But the language's influence is disproportionate to its user base: every modern language with pattern matching, algebraic data types, or type inference carries SML's DNA.

_The failure of Standard ML to conquer industry is not a technical failure but a sociological one. The language demanded that programmers think in types and proofs before they wrote a line of code, and most programmers — even those who would later celebrate the same ideas in Rust — were not ready for that discipline when SML offered it. Standard ML was the right medicine at the wrong time. Its tragedy is that it taught a generation of language designers what was possible, and then watched as they implemented its ideas in languages that compromised just enough to achieve popularity._