Build log

A Test Route Is Not a Test Result

A taxonomy correction: the essays move to /articles/, and the blog address becomes the site’s technical build log.

The false comfort of 200

A route can return a successful response while serving the wrong document. Static hosting fallbacks make this easy to miss: a missing JSON asset can become HTML, and a test URL can appear healthy while its title, heading, canonical address, or page body belongs somewhere else. A status code answers only whether the server responded.

Identity is the real assertion

The site now treats a route check as a small identity test. For a public page, confirm the title, H1, canonical URL, and distinctive content. For a data file, confirm its content type and parse it. For a hub, confirm that the expected card or navigation target is actually present. Those checks catch errors that a broad availability test cannot see.

Why this matters for research tools

The converters are data-backed interfaces. If their data request resolves to a fallback page, the failure shows up in the browser as a parsing error, not as a tidy server error. The repair belongs in the relative path, then in a sibling-route audit, then in a public verification. A reliable calculator is a chain of small correct assumptions.

A reusable release habit

Every release check now distinguishes availability from identity. It is a modest discipline with an outsized payoff: it protects the Study Library, the tools, and the evidence pages from silent fallback behaviour that looks healthy in a superficial report.