What happens before a webpage appears?
After you press Enter, the browser needs an address and a connection to the server. It can then send a request for the page.
The server returns content. The browser still has to interpret its structure, apply styles and arrange it before drawing it on screen. Some work can overlap; some must wait for earlier results.
Loading a page is a relay of different jobs. The film separates them so their roles are clear. Real loading often overlaps tasks and can reuse saved information to avoid waiting.
Why is the second visit often faster?
- Run once, then enable caching and run again. Which waits disappear?
- Increase round-trip delay. Watch repeated trips add up.
- Add a loss and watch reliable delivery repair the gap, with extra waiting.
Separate three ideas
Further controls and numerical checks
Try a cold start, then enable DNS caching. Next, reuse an established connection to skip repeated TCP and TLS handshakes.
Finally, choose fresh resource caching. The browser can reuse responses locally without transferring the same content again. A cached file is not always usable without contacting the server: stale responses may need validation. This experiment specifically models still-fresh resources.
Raise RTT to 300 ms and compare again. Even small transfers can take noticeable time when many round trips accumulate. Avoiding round trips and reducing byte count are different optimizations.
Lose one segment. Reliable transport does not mean nothing ever gets lost; it means loss can be detected and handled.
Want to check the technical details?
After you press Enter, the browser needs an address. A domain is convenient for people; communication needs a network address. DNS connects the two. A usable cached answer makes this step faster.
The experiment uses HTTP/2 over TCP with TLS 1.3. TCP establishes reliable transport. TLS negotiates encryption and verifies the server’s identity. HTTP then expresses the request through that connection.
The server processes the request and returns content. HTML arriving is not the end: the browser parses structure and style, calculates layout, paints and composites the result. Real pages often process content as it arrives.
The animation separates responsibilities. Real loading combines relay-like handoffs with parallel work; it is not always a strictly serial queue.
Bandwidth, latency and rendering
Bandwidth is how much data can travel per unit time. Latency is how long a round trip takes. A wide, long road can carry many trucks without making the first truck arrive immediately.
Server processing, resource count, script execution and device performance also matter. Here processing and rendering times are fixed to highlight round trips and caching. Every readout is a model result, not a measurement of this website or your connection.
HTTP/2 and HTTP/3
HTTP/2 multiplexes streams over a connection, but TCP’s ordered delivery means packet loss can delay later data. HTTP/3 uses QUIC, with different connection establishment and stream behavior. This serial timeline should not be applied directly to it.
DNS can itself be encrypted. Connections can be prepared early and TLS sessions resumed. The sources provide further protocol detail after each responsibility has a clear place in the picture.