Debootstrapping without Archeology

Context: It is common for programming languages that their reference implementation implemented in the language itself. This requires a "bootstrap": copy of previous version provided along with sources, to be able run Those bootstrap files are opaque binaries; they could contain bugs, or even malicious changes reproduce themselves when running source -- this called "trusting trust attack". For reason, collective project Bootstrappable was launched 2016 remove those bootstraps, providing alternative build paths do not rely on binaries. Inquiry: Debootstrapping generally combines mix two approaches. The "archaeological" approach works by locating old versions systems, legacy implementations, need bootstrap, and preserving restoring ability them. "tailored" re-implements new, non-bootstrapped system debootstrap. Currently, dominant low-level components (C, coreutils), among few higher-level were debootstrapped. Approach: We advocate benefits debootstrapping implementations high-level languages. new needs production-ready, it suffices correctly. argue feasible reasonable development effort, several side besides debootstrapping. Knowledge: propose specific design composing/stacking implementations: interpreter interest, small subset language, compiler (in another language). Developing valuable independently debootstrapping: may help clarify semantics, can reused other purposes such as differential testing implementations. Grounding: present Camlboot, our debootstrap OCaml compiler, 4.07. Once we converged final design, last Camlboot took about person-month demonstrating feasibility. Using diverse double-compilation, prove absence trusting attack existing standard implementation. Importance: To knowledge, document first scholarly discussion an interesting problem which recently grew active community free software contributors, but so far interactions programming-language research have been minimal. share experience trying highlight aspects interest designers implementors; hope foster stronger ties between relevant academic communities. In particular, has reflection implementation, implementors would find equally valuable.