What a server is, and writing one in Node

Jem Young28 Aug 20265 min

tl;dr

A server is just a computer that answers requests — there is nothing magical about it. Eight lines of Node prove it, and the interesting part is what those lines say about heads, bodies and streams.

Ask a room what a server is and you get a version of the same answer: a computer that serves requests. That's it. Anything can be a server — your phone, your laptop — if you make it respond to formatted requests.

What we mean by "server" is a set of expectations layered on top:

Dedicated hardware. Server chips cost thousands and use different chipsets, built for virtualization and efficiency rather than for you.
Available 100% of the time. This is the real constraint, and it's the one your laptop fails the moment you close the lid.

Eight lines

const http = require("http")
const fs = require("fs")
const PORT = 3000

const server = http.createServer(function (req, res) {
    res.writeHead(200, { "content-type": "text/html" })
    fs.createReadStream("index.html").pipe(res)
})
server.listen(PORT)
console.log(`Server started on port ${PORT}`)

No npm install. Both http and fs ship with Node.

Every server, in any language — Node, Django, Flask, Rails, Tomcat — is built around the same two objects: a request coming in, and a response going back. Once you see that, framework differences get much smaller.

What actually goes back

A file on disk piped into the body of a response, which also has a head carrying 200 and content-type, then sent to the browser. — Headers aren't part of the body — they're metadata riding alongside it: status, content type, cookies, whether you're logged in.

writeHead(200, ...) sets the status and tells the browser what's coming. Being explicit about content-type is politeness — browsers can often infer it, but guessing is not a plan.

Then createReadStream('index.html').pipe(res). This is the line worth pausing on. Without streams you'd read the whole file into memory, hand it to the server, and only then send it. With a stream you start at the head of the file and pipe it straight through. On a 1 KB file that's irrelevant; on a large one it's the difference between fine and falling over.

Ports 0 to 1023 marked as reserved with SSH on 22, HTTP on 80 and HTTPS on 443; 1024 to 65535 marked as free, with 3000 highlighted. — 3000 is convention, not a rule — nothing binds to it. 8080 and 8000 are equally fine.

Making it run

vi index.html          # "hello world" is enough — browsers are forgiving
brew install node      # Homebrew is the package manager for macOS
node simpleServer.js

Then open localhost:3000.

localhost is a shortcut for 127.0.0.1, a reserved loopback address every computer has — it means look at the open ports on this machine. 192.168.0.1 is another well-known one; it's usually your router.

The HTML file can be a single unclosed line of text and it will still render. Browsers are the most forgiving software on the planet: unknown tags are simply ignored.

And then the catch

You now have a working server that dies when you close your laptop. That's the whole reason the next lesson exists.

SourceFrontend Masters — Full Stack Fundamentals, v3, Jem Young ↗