Gwtar: a static efficient single-file HTML format

2026-02-15 · Source: Simon Willison's Weblog · Field: Technology & Digital — Software Development & Engineering, Emerging Technologies & Innovation · Depth: Intermediate, quick

Summary

Gwtar is a new project by Gwern Branwen and Said Achmiz that addresses the challenge of creating single-file HTML archives containing numerous assets, while maintaining efficient browser viewing. It achieves this by employing a unique mechanism: the page first calls `window.stop()` to halt full document download, then embeds uncompressed tar content inline. Subsequent asset requests are handled via HTTP range requests, fetching data on-demand from the embedded tar. JavaScript rewrites asset URLs to `https://localhost/` to trigger failed loads, which are then intercepted by a `PerformanceObserver`. A `resourceURLStringsHandler` callback retrieves the required resource, either from already loaded data or via a range request, and injects it into the page using a `blob:` URL, ensuring lazy loading of large media files.

Key takeaway

For web developers or content creators needing to package complex web content into a single, portable HTML file, Gwtar offers a novel solution for efficient asset loading. You should consider Gwtar if your goal is to distribute self-contained web pages with large media files without sacrificing browser performance, as it allows for lazy loading directly from the embedded archive.

Key insights

Gwtar enables efficient, lazy-loaded, single-file HTML archives using `window.stop()` and HTTP range requests.

Principles

• Prevent full page download for efficiency.
• Use HTTP range requests for on-demand asset loading.

Method

The method involves calling `window.stop()` early, embedding tar data, rewriting asset URLs to `localhost`, and using a `PerformanceObserver` to intercept failed loads and serve content via range requests or `blob:` URLs.

In practice

• Archive large web pages into a single HTML file.
• Enable lazy loading of media within self-contained files.

Topics

• Gwtar
• Single-file HTML
• Lazy Loading
• HTTP Range Requests
• Self-extracting Archives

Best for: Software Engineer, Research Scientist

Related on AIssential

• — a still that plays — The Web Component defers large GIF loading until user interaction, improving page performance.
• One-Shot Any Web App with Gradio's gr.HTML — Gradio's `gr.HTML` enables single-file, interactive web components with custom HTML, CSS, and JavaScript, ideal for LLM-driven development.
• Building a Pipeline to Study War Coverage Using GDELT — Manual classification can be more efficient and accurate than automation for small, unambiguous datasets.
• DuckLake 1.0: Data Lake Format with SQL Catalog Metadata — DuckLake 1.0 centralizes data lake metadata in a SQL database, improving performance and simplifying operations.
• Google AI Introduces STATIC: A Sparse Matrix Framework Delivering 948x Faster Constrained Decoding for LLM Based Generative Retrieval — STATIC accelerates LLM constrained decoding by transforming prefix trees into sparse matrix operations for hardware efficiency.
• Extract PDF text in your browser with LiteParse for the web — LiteParse for the web enables client-side PDF text extraction using spatial parsing and OCR, built with significant AI assistance.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.