A Web Interface for Matita

This article describes a prototype implementation of a web interface for the Matita proof assistant. The interface supports all basic functionalities of the local Gtk interface, but takes advantage of the markup to enrich the document with several kinds of annotations or active elements. Annotations may have both a presentational/hypertextual nature, aimed to improve the quality of the proof script as a human readable document, or a more semantic nature, aimed to help the system in its processing of the script. The latter kind comprises information automatically generated by the proof assistant during previous compilations, and stored to improve the performance of re-executing expensive operations like disambiguation or automation.

💡 Research Summary

The paper presents a prototype of a web‑based interface for the Matita interactive theorem prover, aiming to bring the full functionality of the traditional GTK client to a browser environment while exploiting HTML markup to enrich proof scripts with both presentational and semantic annotations. Four main motivations drive the work: (1) providing a collaborative, web‑accessible workspace for theorem proving; (2) using markup to embed hyperlinks, disambiguation hints, and automation traces directly in the source; (3) achieving platform independence and low‑resource accessibility; and (4) removing installation barriers that typically deter new users, especially in the mathematical community.

The authors identify three categories of markup supported by the system. Standard HTML markup is used for visual formatting inside Matita comments. Hyperlinks to Matita definitions are generated automatically during parsing; they encode disambiguation information so that the prover does not need to ask the user to resolve ambiguous identifiers on subsequent runs. Finally, traces of automation tactics—produced on the first execution—are stored as hidden markup that can be displayed on demand, dramatically speeding up later executions of the same script.

On the server side a minimally modified Matita engine runs inside an Ocamlnet/Netplex CGI daemon. The engine’s state now includes a user identifier, enabling multiple concurrent sessions each with its own library version. The lexical analyzer ignores markup tokens, while a side table records hyperlink information that the parser later consults to build a disambiguated abstract syntax tree (AST). This design preserves the existing parser’s complexity, assuming that semantic markup appears only around leaf nodes (identifiers or symbols), an assumption that currently holds for the prototype. Automation‑trace markup is handled entirely by the UI and therefore does not affect the lexer or parser.

The daemon offers several services via XML‑encoded CGI responses: user account management, storage of personal libraries, synchronization with a central SVN repository, and remote script execution. Because Matita’s syntax is extensible at runtime, client‑side parsing is infeasible; the server must receive the whole remaining script fragment, parse it, and return (i) the length of the portion successfully executed, (ii) a possibly empty list of newly parsed statements enriched with generated markup, (iii) an HTML rendering of the current proof state (including open goals), and (iv) any error information that halted execution.

Collaboration is currently realized through a centralized SVN repository, reusing Matita’s original code base. While other projects have advocated distributed version control systems such as Git, the authors chose SVN for its compatibility with the existing Matita workflow and its support for the modest distributed activity envisioned. Future work will evaluate scalability and may explore more decentralized approaches.

The client is written in plain JavaScript, now being ported to the jQuery framework. Its user interface mirrors that of Proof‑General, CoqIDE, and the standalone Matita UI: the script is displayed with the already‑executed part locked, navigation buttons allow step‑wise or whole‑script execution, goal windows show multiple open goals, disambiguation panels appear when needed, and TeX‑like escapes are converted to Unicode on the fly. The client does not need deep knowledge of Matita internals; it simply interprets the XML data returned by the server, converting Matita‑specific markup to HTML and vice‑versa where necessary.

The prototype is publicly accessible (login provided on request). The authors argue that the web interface eliminates the need for local installation, leverages the browser’s native rendering capabilities for rich, hyperlinked documents, and reuses expensive disambiguation and automation results through markup, thereby improving performance. Limitations include the current lack of on‑the‑fly hyperlink traversal (the server must re‑parse the entire script for each step) and the reliance on the “leaf‑only” markup assumption, which may restrict future extensions. Moreover, SVN‑based collaboration may not scale as well as distributed systems for large, geographically dispersed teams.

In conclusion, the paper demonstrates that a web‑based front‑end for an interactive theorem prover can retain the expressive power of the underlying system while providing a more accessible, collaborative, and performance‑optimized environment. The authors plan to extend the prototype with real‑time hyperlink‑driven re‑parsing, richer semantic markup, and integration with distributed version control, paving the way for broader adoption in both educational and research settings.