A Certified JavaScript Interpreter

The JavaScript language was initially developed for web pages enrichment, allowing the execution of scripts by the browser. It is now pervasively used on the web, not only to add interactivity in websites or to embed contents from third-party sources, but also as a target platform for deploying applications written in other languages (such as ocaml bytecode [VB11], Hop [SGL06], or LLVM assembly [Zak11]). In some sense, JavaScript has become the assembly language of the web, as most browsers are now able to run it. More recently, it has been used to program user interfaces for embedded systems, such as the defunct WebOS (now Enyo [Eny12]), the Kindle Touch ebook reader, or for the BootToGecko project [Moz12]. In addition to its pervasive use, JavaScript presents two important characteristics. First, as it was initially developed to facilitate its integration with the browser and with web contents, it aims more at providing powerful features than at giving robustness and safety guarantees. These powerful features include first class functions and closures, prototype-based objects, dynamic typing with many conversion functions, explicit scope manipulation, and the evaluation of strings as code. A second, redeeming, characteristic of JavaScript is that it is standardized [A99], providing more information about how these features interact. The goal of the JSCert project [BCF12] is to provide a precise and formal semantics to JavaScript to build tools to certify analyses and compilation procedures. JSCert’s collaborators have defined such a semantics in the Coq proof assistant, based both on the paper formalization of Maffeis et al. [MMT11, MMT08] and on the specification. To gain and provide more confidence in this formalization, we have implemented an interpreter that is proven correct in relation to the semantics. We will thus be able to confront our semantics against JavaScript test suites. This paper describes the design and implementation of the interpreter. It is organized as follows. Section 1 introduces the semantics of JavaScript and highlights some of its peculiarities. Section 2 describes the interpreter’s design and implementation. Section 3 addresses the interpreter’s correctness. Finally, Section 4 concludes with future and related work.