How this site is built

We argue that organizations should not hand their most valuable data to a third party and hope the contract holds. Making that argument while running a careless website would be unpersuasive, so here is our own homework.

One thing to say plainly first. This is a marketing site. It holds nothing commercially sensitive, so it is the easy version of the problem. Anyone can run a few DNS queries and see that it is served from a cloud provider rather than a rack we own. What it does demonstrate is the discipline, and the discipline is what carries over to work that does matter.

It is written in React with MUI components and turned into a static HTML file for each page before it is published. There is no application server and no database behind it. It is served from object storage behind a load balancer.

Because each page is already HTML, no JavaScript is required to read it. Scripts add the menus, the analytics choice and the contact form.

The fonts are ours. They are served from the same place as the page, so no font provider gets a record of your visit.

Everywhere but the contact page, one third-party request exists, and it is optional. Analytics does not load unless you accept it. Decline, or simply ignore the banner, and nothing is requested and no cookies are set. The contact page is the one exception: its form sends to our own support desk and uses Google reCAPTCHA to keep out automated spam, which loads from Google and sets Google's cookies on that page only. Both are in the privacy policy and the cookies policy.

A missing page returns a real 404, not a 200 with an apology on it.

Every change to this site is an issue, a branch, a pull request and a cryptographically signed commit. Merging to the main branch builds the site and publishes it, so what you see is what is in version control.

No deployment credential is stored anywhere. The build proves its identity with a short-lived token issued for that single job, exchanged for temporary cloud credentials that expire on their own.

Plenty. A static site with no user accounts and no uptime obligation is a soft target for a page like this. It does show what we consider normal: check the thing rather than assume it, write down the result, and do not quietly leak your visitors to third parties because it was the default.