Koka is a research-oriented functional programming language developed by Daan Leijen and his team at Microsoft Research. It is a beautifully designed language that introduced the world to the modern, practical application of algebraic effects, a concept that is a cornerstone of Ribbon’s entire design. Our debt to the profound insights and academic rigor of the Koka project is immense.

The Power of Algebraic Effects

Koka’s primary innovation is its first-class support for algebraic effects. It provides a clean, elegant, and powerful way to separate pure, logical code from the messy, side-effect-ful code it needs to interact with. An effect system allows a function to be defined purely in terms of the what (its logic) while deferring the how (the implementation of its side effects) to a caller-provided handler.

This makes code incredibly modular, highly testable, and easy to reason about. The ability to swap in different effect handlers for testing, or to compose complex behaviors from simple handlers, is a paradigm shift for software architecture. Nearly everything Ribbon has to say about the power of Algebraic Effects was said first and with more academic precision by the Koka team.

A Different Foundation

Where Ribbon’s philosophy diverges from Koka’s is at the most fundamental level: the memory model and runtime.

Koka is a purely functional language that provides automatic memory management. It uses a state-of-the-art system of reference counting combined with its novel Perceus algorithm for compile-time reuse analysis. This is a brilliant approach that makes memory management both efficient and highly predictable within a functional paradigm, eliminating entire classes of programmer concerns.

However, for Ribbon, whose roots are in imperative, performance-focused game development, any form of automatic, non-deterministic memory management, even one as advanced as Perceus; was a trade-off we couldn’t make. The core requirement for a true systems language is giving the programmer explicit, deterministic control over the layout and lifetime of memory.

• Koka’s Goal: Abstract away memory management to provide the programmer with a pure, safe, and highly productive environment.

• Ribbon’s Goal: Expose memory management as a first-class, controllable part of the system, via Composable Allocation Strategies and arena-based allocation.

Koka’s Interface, a Systems Implementation

Ribbon’s design is a synthesis of these two worldviews. We fell in love with Koka’s beautiful interface for effects but knew we needed a different implementation to meet our performance goals.

The result is that Ribbon’s effect system, while conceptually identical to Koka’s, is implemented in a way that is familiar to a systems programmer. It is built on a foundation of stack unwinding and direct jumps, mechanically similar to a highly-optimized “typed exception” system. This provides the powerful separation-of-concerns benefits pioneered by Koka, but with a runtime model that has zero overhead on the “happy path” and a predictable, low-level execution model that is suitable for the most performance-critical applications.