HTTP Caching

HTTP Caching is the family of headers that let a browser, a CDN, or a proxy skip a request entirely, or skip its body, by reusing a response it already has. Cache-Control is the modern header doing most of the work: max-age sets how long a response is fresh without any revalidation at all, no-cache means always revalidate before using it (despite the name, it doesn't forbid caching), and no-store means never cache it at all — three genuinely different behaviors that a name like "no-cache" makes easy to mix up.

Once a cached response goes stale, the client doesn't necessarily re-download it — it can revalidate: ask the server "has this changed?" with a conditional request carrying an ETag or a last-modified date, and the server answers 304 Not Modified with no body if nothing changed, saving the bandwidth of the full response while still confirming freshness. This is the mechanism that makes aggressive caching safe: a max-age that expired doesn't mean a wasted round trip, just a cheap one.

Static-asset caching pushes this further by making the URL itself change whenever content does — a hash embedded in the filename (app.a3f9c1.js) — so the response can be cached with an effectively infinite max-age and no revalidation ever needed; a new deploy simply produces a new URL. That trick is what makes Preload and Prefetch and long-lived CDN caching viable together: the browser never has to ask whether a hashed-filename asset is stale, because a stale one is, by construction, a different file.

This layer is orthogonal to Cache Invalidation inside an application's own data layer — HTTP caching governs whether a response gets reused, application caching governs whether computed data does, and the two commonly disagree about what's fresh at the same moment.