Every request you make to a website carries a small pile of metadata you never see. Headers. They decide whether your connection is encrypted, whether a page can be embedded in an iframe, which CDN edge served you, and whether the browser should remember a cookie for a year. I wanted to see what a real, busy production site sends, so I pointed curl at an Amazon endpoint and dumped the response headers. Turns out there’s a lot to unpack.
Amazon didn’t set out to build the world’s largest cloud platform. It stumbled into one while trying to stop its own engineers from reinventing the same infrastructure wheel every few months.
The Internal Mess That Started It All Around 2000, Amazon was building Merchant.com — a product to let third-party retailers like Target and Marks & Spencer spin up e-commerce stores on Amazon’s infrastructure [2]. What followed was an organizational disaster. Every team was independently building its own version of the same storage, compute, and database primitives. No shared APIs, no standard way to access anything.