The smallest (useful) HTTP responses possible

I recently did a little experiment to see how small (in bytes) I could make useful valid HTTP responses for a particular use case.

Background

On my homepage, I have a little feature where I display the names of the last song I listened to and the last movie I watched. I’ve implemented that via a bit of JavaScript that fetches that information from a pair of Lambda functions that make requests to Last.fm and Letterboxd.

I use that client script as a place to play some code golf with myself, trying to achieve the functionality with as few JavaScript tokens as possible.1 Every few months, I think of a way to shrink it further. Here’s what I’ve got as of the writing of this post (source):

...listening to music<span id="song"></span>,
watching movies<span id="movie"></span>...

<script>
  [song, movie].forEach(async (element) => {
    const response = await fetch(`/.netlify/functions/${element.id}`);
    if (response.ok) element.innerHTML = ` (${(await response.text()).trim()})`;
  });
</script>

The responses from the Lambda functions (source) look like this:

the last song I listened to was
<a href="https://www.last.fm/user/magicjamesv">
  Nonagon Infinity by King Gizzard & The Lizard Wizard
</a>

This makes it trivial to shove them into the page without having to parse or modify them.

Going smaller

That script is pretty small in terms of JavaScript tokens, but we can go smaller! As of this post, I have two ideas:

Combine the Lambda functions

I could combine the two Lambda functions into a single one that returns markup for replacing the entire paragraph.

With that change, the client script look something like this:

<p id="paragraph"><!-- The whole original paragraph --></p>

<script>
  const response = await fetch(`/.netlify/functions/combined`);
  if (response.ok) paragraph.innerHTML = await response.text();
</script>

I don’t really want to do this for a couple reasons:

  1. I don’t want to move or copy the main text of that paragraph into the Lambda function.
  2. I want this to continue being able to half-work. With my current implementation, if something goes wrong with one of the requests, the other can still work.

Remove the second async call

The script currently has two async calls: fetch() and response.text(). The first one resolves once the HTTP response’s headers have been received and the second resolves once the body has been received.

The fetch() will be difficult-to-impossible to remove, but the response.text() one is a bit more avoidable! My first thought was to put my response data in an HTTP response header, but then I realized I could go further and extend my code golfing to the HTTP response itself, putting the data directly into the status text.

After some experimentation and iteration, I got this minimal demo working:

import { createServer } from "net";

const data = "Hello World";

createServer((socket) => {
  socket.write(`HTTP/1.0 204 ${data}\r\n\r\n`);
}).listen(3003);

There are a few things going on there:

  1. I’m using createServer and writing to a socket so that I can have more control over the bytes in the response.
  2. I’m using HTTP 1.0 because clients automatically close the connection after two consecutive CRLFs without me having to explicitly send a Connection: close header like I’d need to with HTTP 1.1 or newer.
  3. I’m using a 204 response code (the code for a No Content response) because clients won’t try to wait for a response body when they see that code.

Here’s how that looks when fetched via cURL:

$ curl -v "[::1]:3003"
*   Trying [::1]:3003...
* Connected to ::1 (::1) port 3003
> GET / HTTP/1.1
> Host: [::1]:3003
> User-Agent: curl/8.7.1
> Accept: */*
>
* Request completely sent off
* HTTP 1.0, assume close after body
< HTTP/1.0 204 Hello World
<
* Closing connection

On the browser side, here’s how I could make use of a response like this in my homepage script:

[song, movie].forEach(async (element) => {
  const { ok, statusText } = await fetch(`/.netlify/functions/${element.id}`);
  if (ok) element.innerHTML = ` (${statusText})`;
});

That’s smaller!

I’m not actually using this technique on my homepage because I haven’t found a way to return HTTP 1.0 responses from Netlify Functions. I could set up my own server to run this, but that’s not quite worth it to me for this code shrinkage.

Complete demo

Here’s a more complete demonstration of this technique in a single JS file:

// statusTextDemo.js
import { createServer } from "net";

createServer((socket) => {
  socket.on("data", (data) => {
    const request = data.toString();
    const [requestLine] = request.split("\r\n");
    const [_method, path] = requestLine.split(" ");

    if (path === "/") {
      const body = `
        <h1>HTTP response text demo</h1>
        <div id="result">Loading...</div>
        <script>
          fetch("/data").then(r => result.innerHTML = r.statusText)
        </script>
      `;
      socket.write("HTTP/1.0 200 OK\r\n");
      socket.write("Content-Type: text/html\r\n");
      socket.write(`Content-Length: ${body.length}\r\n`);
      socket.write("\r\n");
      socket.write(body);
    } else if (path === "/data") {
      const responseData = "It worked!";
      socket.write(`HTTP/1.0 204 ${responseData}\r\n\r\n`);
    } else {
      socket.write("HTTP/1.0 404 Not Found\r\n\r\n");
    }

    socket.end();
  });
}).listen(3003);

Copy that to a file, run it via your favorite JavaScript runtime, and open localhost:3003 in your browser. If you see “It worked!” then it worked!

Going even smaller

There are at least a few ways that useful HTTP responses can be even smaller than this:

  1. Responses where the status code itself is the useful data. In the right context, HTTP 1.0 XXX as the entire response (where XXX is any response code) can be useful.
  2. HTTP 0.9 Simple-Response, where no status line is required.

I’m not using those in these experiments because I do need more information than just the status code and because browsers have largely removed support for HTTP 0.9 responses.


  1. Don’t worry, this is just a fun challenge for myself and is not how I typically code. ↩︎

Discuss this post here on Hacker News or shoot me an email!