Re:Proof that any fixed-axis type system fails for some doma...
by trissim on Tuesday January 06, 2026 @12:54AM
(#46509174)
I formalized a proof that the structural vs. nominal typing debate can't be won: any fixed-axis type system necessarily fails for some domain.
The core result: for a type system with fixed axes (like behavior, structure, hierarchy), there exists a domain whose requirements cannot be fully captured. This isn't a limitation of specific languages, it's a mathematical impossibility result.
What's in the zip
Full Lean 4 formalization (compiles with lake build)
Instantiations for Python, TypeScript, Java, and Rust
The axioms for type system structure are extracted from actual language specifications, not invented
Main theorem: For any fixed-axis type framework, there exists a domain D such that no complete typing is achievable.
The core result: for a type system with fixed axes (like behavior, structure, hierarchy), there exists a domain whose requirements cannot be fully captured. This isn't a limitation of specific languages, it's a mathematical impossibility result.
What's in the zip
Full Lean 4 formalization (compiles with lake build) Instantiations for Python, TypeScript, Java, and Rust The axioms for type system structure are extracted from actual language specifications, not invented
Main theorem: For any fixed-axis type framework, there exists a domain D such that no complete typing is achievable.