HTTP Server in Java — From Scratch

Repo: github.com/soufian-elouazzani/http-server

Context

Frameworks like Spring Boot and servlet containers hide the mechanics of HTTP. Before relying on them in production, I wanted to understand exactly what happens on the wire when a browser sends a request and a server responds.

The goal: a minimal but correct HTTP server in plain Java that accepts connections, parses requests, routes to handlers, and writes proper responses — no external web framework.

What I built

• A TCP server that accepts connections on a configurable port and reads byte streams.
• An HTTP request parser that extracts the method, path, headers, and body from raw input.
• A router that maps paths and methods to handler functions.
• Response builders that write status lines, headers, and bodies back to the client.
• Support for common status codes (200, 404, 500) and static file serving.

Technical decisions

• Blocking I/O first, optimize later — A thread-per-connection model is simple and correct for learning. It makes connection lifecycle and resource cleanup visible.
• Parse before route — Separating parsing from routing means malformed requests fail fast with 400 responses instead of crashing handlers.
• No framework magic — Every line of code is traceable. When something breaks behind nginx or a load balancer later, I know what the origin server is actually doing.

Results and takeaways

• The server handles concurrent browser requests and serves static HTML/CSS correctly.
• Debugging taught me why keep-alive, content-length, and chunked encoding matter — details frameworks handle silently until they do not.
• For recruiters and clients: understanding HTTP at this level means faster debugging when APIs misbehave, CORS blocks requests, or proxies strip headers.

Stack

Links

• github.com/soufian-elouazzani/http-server