Why this site is a folder of files

21 October 2026 · Bristol

Four HTML files, one stylesheet, a sitemap. It has been up for two years and the only maintenance it has needed was renewing a certificate the CDN handles for me. I have run enough things on servers to be confident that this is not laziness.

What a static site removes

No runtime means no dependency upgrade treadmill and no process that can be alive but not serving. The failure modes reduce to DNS, TLS and whether the file exists. Two of those belong to the CDN. The third is a directory listing.

There is also the boring benefit: deploys are atomic. Upload the new files, check two pages, done. Rollback is the previous upload, not a migration in reverse.

Where it stops working

The moment a page needs per-user state, you are building a client application and an API, and the complexity has moved rather than disappeared. Search is the other wall. A prebuilt index works to a few thousand documents and then it does not.

I am not arguing that everyone should do this. I am arguing that a lot of sites which are currently containers exist because the team already had a container platform, not because the site needed one.

Small things that matter more than they look

Cache HTML briefly and assets forever. The reverse is the classic mistake and it freezes the site for returning visitors.

Serve a real 404. A soft 404 that returns 200 with an apology is a gift to nobody, and crawlers handle it about as well as users do.

Do not block crawlers by default. I have seen three sites this year that had a `Disallow: /` left in from a staging deploy, one of them for eleven months, and nobody noticed because the site "worked fine".

Put security headers at the edge. They belong to the response, and the edge is where every response passes.