Fixture entry 1: a page built from a file

This entry exists to prove one thing: that a page added as a plain file is published everywhere it should be, without anyone touching the HTML of the site. The file you are reading was written into the articles folder with a small header and a body in markdown. Nothing else was done by hand. The build read the header, placed this page under its section, put its card at the top of the home page because its date is the most recent, wrote its line in the sitemap, added its item to the feed and served its photograph in four widths. If any of those surfaces were missing, the build would have refused the whole package rather than publish a half page.

What does the header decide?

The header carries the title, the date, the section, the commercial category, a description, the scene of the photograph and, when the entry cites another site, the address of that site and the words of the anchor. The date is not decoration. It orders the home page, it feeds the sitemap and it dates the item in the feed. The section decides the first segment of the address and the reference number of the entry. The category is a slug from a shared list of twenty-five, which the backend copies into its own table when the page is published, so that the site and the database never disagree about what the page is about.

Where does the link go?

When a target address is present, the engine looks for the anchor in the first third of the body. A plain description is the clearest way to see it: the sentence you are reading carries the anchor, and the engine wrapped it in a link because it sits early enough in the text. If the anchor had been further down, the engine would have inserted the phrase given in the header after the first paragraph. If neither option worked, the build would have stopped with a message naming the file. There is never a second link to the same site on the page, and there is never a block of links at the bottom that announces where the reader is being sent.

How is the home page rebuilt?

The home page is generated from the same data every time. The most recent entries appear first, each with its photograph, its reference number, its section and its date. Below them, each section lists its latest entries. At the bottom, before the footer, a section reads the file of outbound links and renders one item per line of that file, in the words written for it. Adding a line to that file and rebuilding is all it takes to add a link to the home page; removing the line removes the link. There is no template to edit and no worker to redeploy.

What does the worker do with an unknown address?

The worker serves any file that exists in the package first. Only when the package has nothing at that address does it look at the table of historical addresses, which can answer with a permanent redirect or a gone status. Everything else receives a not-found page with a noindex header. A new page therefore never has to be declared to the worker: the file exists, so it is served.

What is checked before anything is published?

The build counts what is missing and prints it: a card without a photograph, a link outside the first third, a page that no other page links to, a page missing from the sitemap, an image without width and height, a dash of the wrong kind, a phrase that admits to a network. Any of these stops the build. After deployment, the probes read the live site: the page answers, the link is unique and early, the card is first on the home page, the sitemap line and the feed item exist, the photograph is served, an unknown address answers not found, and the www host redirects to the bare domain.

How is this entry removed?

The file is deleted, its photograph is deleted, the site is rebuilt. The page, its card, its sitemap line, its feed item and its derived images disappear together, because none of them was ever written by hand. The fingerprint of the source folder returns to what it was before the test, and that comparison is part of the test.