Mission

Fetch example.com/index.html and count how many requests one page really takes.

http request / response · over one TCP connection0 bytes of HTTP

raw bytes on the wire · plain text, human readable—

connection open, nothing asked yet — press send below

http_message · a start line, some headers, then a body

method: GET
path: /index.html
host: example.com
status: —
content-type: —
content-length: —
etag: —
connection: —

exchange_state

Visit

idle · socket open

1TCP connection already open

2GET /index.html — the request

3Server resolves path to a resource

4200 OK + headers

5Body streams — 13,412 bytes

63 more requests — same connection

Think first · step 1 of 6

HTTP can’t move bytes on its own. What has to exist before the browser can send anything?

what just happened

A GET request is a few lines of plain text. Step through and watch one request turn into one response — and then watch how one page turns out to need four of them.

socket 49512 → 203.0.113.20:80 ESTABLISHED 0 bytes exchanged

the common mistake

One page is not one request. The HTML arrives first, and only then does the browser discover the CSS, the script and the image it needs — each one is its own request over the same connection.

HTTP is stateless: the server remembers nothing between requests. Every request has to carry its own context — that is why Host, cookies and auth headers are repeated every single time.

HTTP does no delivery work of its own. It rides a TCP connection that was already built by a three-way handshake, to an address found by DNS.