DNS, Cloudflare Workers and Framer!

DNS, Cloudflare Workers and Framer!

Everything on the internet starts with DNS 🌐, so it’s kinda important ⭐. Cloudflare is one of the top DNS providers and it offers the unique capability of intercepting any HTTP(S) request right after DNS resolution and mess around with it 🚀. Let’s see how this works and how we have used it effectively at Invisible to handle elegantly a thorny issue with a 3rd party provider.

“It’s not DNS. There’s no way it’s DNS. It was DNS” ~ Every network engineer, ever

🖼️ Bring your own domain to Framer 🖼️

We use Framer to design, build and deploy landing pages and static pages for products like Taka. It’s great for non-engineers to design and create rich web sites. Our setup is to manage the static parts in Framer, where our designers can iterate and evolve the design without support from engineers and then the dynamic part is a React/typescript web application. Note that Framer is not just a design tool, it hosts the websites you design.

This works great for clear division of responsibilities between designers and engineers, but there is this sticky part of accessing the Framer sites from your browser. Framer provides for each project (website) an auto-generated URL like https://weird-name-123456.framer.app. There is also an option for a free custom sub-domain like <your domain>.framer.website.

We obviously want our users to access our websites and products through our own domain names. Now, Framer allows connecting your own 3rd party domain. The configuration is pretty elaborate (see https://www.framer.com/help/articles/how-to-connect-a-custom-domain) and we had to handle multiple auxiliary issues like redirecting from HTTP to HTTPS and from the root domain to www using special rules on the Cloudflare side. It took awhile to figure it out, but eventually we got it under control. We had a well-documented process to configure Cloudflare and Framer with our own domains. Life was good 😊.

🚫 Framer and Clickjacking 🚫

Then we discovered that Framer is susceptible to clickjacking attacks. Clickjacking 🎭 is a malicious technique to lure users to click on things they’re not supposed to. I won’t get into too much detail here. The important thing is that you protect against Clickjacking by adding some HTTP headers to every HTTP response. Easy peasy 🫛, right?. Well, guess what? Framer doesn’t do it and doesn’t want to do it. Their answer? On static pages it can’t do much harm and if you care so much just use a reverse proxy.

There is an old community thread with no response from Framer - kind of ironic.Isn’t it?

Well, that was unacceptable to us. But, we didn’t want to drop Framer, so we had to come up with a solution. Enter Cloudflare workers!

👷 Cloudflare Workers 👷

Cloudflare Workers are a general-purpose serverless platform that runs lightweight code at Cloudflare’s edge network, enabling you to deploy applications, APIs, and custom logic closer to users for low-latency performance and scalability without managing extra infrastructure (e.g. no need to deploy a reverse proxy server).

For our purpose, they can intercept any HTTP(S) request immediately after DNS resolution, allowing you to modify requests, route or redirect them, and alter responses before they are sent back to the client.

Cloudflare is unique in this respect. Other CDN providers offer similar functionality, but only for requests to CDN content after DNS resolution is over.

🦜 Killing Three Birds with one Cloudflare Worker 🦜

I hope I don’t get in trouble for using the expression “Killing three birds…” :-)

Anyway, I built a Cloudflare worker called “framer-gateway-worker”. It performs three functions:

  1. Routes from our URLs directly to Framer URLs (users never see the Framer URLs)
  2. Update page metadata URLs
  3. Add the proper HTTP headers for Clickjack protection before returning the responses from Framer to the users.

Let’s see how it works. The Cloudflare worker code is a simple Javascript program that listens for the “fetch” event and calls the handleRequest() function.

addEventListener('fetch', event => {
  event.respondWith(handleRequest(event.request))
})

It also has a map for translating from our domain names to the corresponding Framer domain names

const hostnameMap = new Map([
  ['getinvisible.com', 'broader-memory-661481.framer.app'],
  ['www.getinvisible.com', 'broader-memory-661481.framer.app'],
  ['staging.takafinance.com', 'rejuvenated-pentagon-519952.framer.app'],
  ['takafinance.com', 'takafinance.framer.website'],
  ['www.takafinance.com', 'takafinance.framer.website']
])

The handleRequest() function first modifies the request target URL by replacing the original [sub-]domain with the Framer sub-domain from our hostname map,

async function handleRequest(request) {
  const url = new URL(request.url)
  if (hostnameMap.has(url.hostname)) {
    url.hostname = hostnameMap.get(url.hostname)
  }

  const newRequest = new Request(url.toString(), request)
  const response = await fetch(newRequest)

Once the response is received, it replaces mentions of framer.website with .com. For example, takafinance.framer.website will become takafinance.com. This is important because the HTML head section contains metadata nad canonical URLs that need to be fixed

let html = await response.text();

html = html.replace(/\.framer\.website/g, ".com");

Then, it adds the Clickjack protection headers and returns the modified response

const modifiedResponse = new Response(html, response)
modifiedResponse.headers.set('Content-Security-Policy', "frame-ancestors 'none';")
modifiedResponse.headers.set('X-Frame-Options', 'DENY')
return modifiedResponse
}

Now, we don’t want to run this for EVERY request. The idea is to limit it to pages with specific domain names that will require either the routing or the clickjack protection. We can configure this using Cloudflare routes in the wrangler.toml file:

name = "framer-gateway-worker"
main = "index.js"
compatibility_date = "2022-11-30"
account_id = "<redacted>"

workers_dev = false

# wrangler.toml (wrangler v3.79.0^)
[observability]
enabled = true
head_sampling_rate = 1

[env.production]
name = "framer-gateway-worker"
routes = [
  { pattern = "getinvisible.com/*", zone_name = "getinvisible.com" },
  { pattern = "www.getinvisible.com/*", zone_name = "getinvisible.com" },
  { pattern = "takafinance.com/*", zone_name = "takafinance.com" },
  { pattern = "www.takafinance.com/*", zone_name = "takafinance.com" },
  { pattern = "staging.takafinance.com/*", zone_name = "takafinance.com" }
]

Note that we still have to create DNS A records proxied to Cloudflare (AKA Orange cloud) in order for Cloudflare to try to resolve DNS requests for these domains and then because the routes exist it delegates the heavy lifting the Cloudflare worker.

The content of the DNS A records doesn’t matter because the worker takes care of it, but we must choose some value even if it is ignored. The best practice for placeholder IP addresses is to use an IP addresses from one of the documentation blocks. See https://www.rfc-editor.org/rfc/rfc5737.

I chose 192.0.1, which is very common.

🏡 Take home points 🏡

  • Framer is cool, but its custom DNS game is weak
  • Cloudflare workers can sidestep Framer’s limitations
  • I barely scratched the surface regarding Cloudflare workers

💡 Cloudflare workers are the eBPF of the Internet.💡

Buon fine settimana, miei amici! 🇮🇹