6 questions
Questions, answered plainly
The awkward ones first. If something here is still unclear, the answer is a reply away rather than a form away.
- Why are unfinished projects listed at all?
- Because leaving them out would make this a portfolio rather than a record. Not every entry has been deployed, and the register says so on each one that has not. The interesting thing about a body of work is its actual shape, which includes what stalled and why, and a page of nothing but successes tells you nothing about how someone works.
- Why do some entries have no link to a live site?
- Because they are not deployed, and an entry with no deployment shows no link. Four projects are reachable; three run but have never been published; one has nothing to visit yet. Where a URL is genuinely unknown the field is empty rather than filled with a guess.
- Where does the register's data actually live?
- In a Postgres database on one machine, which has no public network path — not even a tunnel. Editing happens there. The published site is a static snapshot generated from that database and committed to the repository, so the version you are reading was built from the data rather than querying it.
- What happens to this site when that machine is asleep?
- Nothing. That is the reason for the snapshot. The site is static and served from a CDN, so it is up whether or not the machine that holds the database is running. The only thing that needs the machine awake is adding or editing an entry.
- How does a new project get added?
- One command inserts a validated row and regenerates the snapshot, and the commit is the deploy. The accession number is assigned by the database rather than typed, so it cannot be duplicated or chosen, and it never changes once issued. That is what makes it an accession number rather than a row id.
- What is sites.json for?
- It is the register as machine-readable data, so an assistant or a script can read it without scraping the page. It carries the same entries you can see, minus the parts that are nobody's business — filesystem paths stay on the machine and never appear in the response.
Every message gets a reply within two working days. Nothing here is handled by an autoresponder.