RFC 9111: Retiring Wasteful Pre-Render Solutions

Paying thousands of dollars a month to run headless Chromium clusters that pre-render JavaScript into disposable HTML snapshots for bots is an architectural tax on a solved problem. RFC 9111 Section 4.1 (Calculating Secondary Keys with Vary) lets your CDN cache both human HTML and agent-ready Markdown at the network edge from one origin URL.

Inspect the reference harness in less first (scroll with j and k, quit with q):

curl -fsSL https://rfc9111.io | less

When you are ready to launch the local edge-caching audit workbench in ~/rfc9111:

curl -fsSL https://rfc9111.io | bash

Why Bimodal Edge Caching Beats Dynamic Pre-Rendering

• Zero Headless-Browser Compute Tax — static HTML and static Markdown sit side-by-side in edge RAM, partitioned cleanly by Vary: Accept, User-Agent.
• Zero Cloaking Liability — every representation is declared openly in standard HTTP response headers rather than hiding behind brittle user-agent sniffing middleware.
• Instant Time-to-First-Byte for AI Grounding — RAG and search agents receive compact Markdown in milliseconds instead of waiting for a headless browser spin-up.

The book this grew out of · RFC 9110 Semantics · Sun Pug · QA My AI · NPvg · Levinix · GuerillaTech · Pipulate