How It Works

Overview

When you run falcon serve, Falcon creates a Falcon::Controller::Serve which is used to create several worker threads or processes. Before starting the workers, the controller binds to an endpoint (e.g. a local unix socket, a TCP network socket, etc). The workers are spawned and receive this bound endpoint, and start accepting connections.

The workers individually load a copy of your rack application. These applications are wrapped using Falcon::Adapters::Rack which modifies the incoming Protocol::HTTP::Request object into an env object suitable for your application. It also handles converting the output of your rack application [status, headers, body] into an instance of Falcon::Adapters::Response which is derived from Protocol::HTTP::Response.

See the protocol-http documentation for more details on how it works.

Server

The server itself is mostly implemented by Async::HTTP::Server which in turn depends on the protocol-http gems for the actual protocol implementations. Therefore, Falcon is primarily a bridge between the underlying protocol objects and the Rack interface.

See the async-http documentation for more details on how it works.

Protocol::Rack

Falcon uses the protocol-rack gem to provide a Rack interface for the HTTP protocol. This allows you to run any Rack-compatible application with Falcon. However, Falcon itself is not a Rack server, but rather an HTTP server that can run Rack applications.

See the protocol-rack documentation for more details on how it works.