The site must not advertise a feature that does not work #8

New issue

Open

opened 2026-08-24 23:16:56 +00:00 by jlxq0 · 0 comments

jlxq0 commented

2026-08-24 23:16:56 +00:00

Owner

Filed so the intent survives the mechanism. The mechanism is being removed deliberately — EltrixSite.Status.claim!/1 and the GOAL.md artefact behind it. This is not a request to put it back.

What was guaranteed, and no longer is

The site's build failed when a page claimed a capability whose status was below done. The moduledoc put it well: "the site never asks a person whether a feature works, it asks this file."

That existed because of a specific failure, not a hypothetical one. The previous site advertised an Admin API, a Helm chart, Redis caching and read-replica routing. None of it was malicious. Somebody wrote a roadmap in the present tense and nothing anywhere could tell.

The second half mattered as much: the same page rendered what was partial and what was missing, from the same source, with no way to render one list and quietly drop the other. A site that lists only finished features is not lying by sentence and is lying by shape.

If a replacement is ever wanted

It reads from Forgejo, not from a file. Issue and milestone state is the fleet's record of what is done; a file is not, which is the reason the file went. A build-time query against the v1.0 milestone would give the same guarantee with the same single source and no document to keep in step.

Until then this is a discipline rather than a check, and discipline in a repository with a public marketing surface has a known half-life. That is the trade, and it is recorded here rather than discovered later.

Filed so the intent survives the mechanism. **The mechanism is being removed deliberately** — `EltrixSite.Status.claim!/1` and the `GOAL.md` artefact behind it. This is not a request to put it back. ## What was guaranteed, and no longer is The site's build failed when a page claimed a capability whose status was below `done`. The moduledoc put it well: *"the site never asks a person whether a feature works, it asks this file."* That existed because of a specific failure, not a hypothetical one. The previous site advertised an Admin API, a Helm chart, Redis caching and read-replica routing. **None of it was malicious.** Somebody wrote a roadmap in the present tense and nothing anywhere could tell. The second half mattered as much: the same page rendered what was **partial** and what was **missing**, from the same source, with no way to render one list and quietly drop the other. A site that lists only finished features is not lying by sentence and is lying by shape. ## If a replacement is ever wanted **It reads from Forgejo, not from a file.** Issue and milestone state is the fleet's record of what is done; a file is not, which is the reason the file went. A build-time query against the `v1.0` milestone would give the same guarantee with the same single source and no document to keep in step. Until then this is a discipline rather than a check, and discipline in a repository with a public marketing surface has a known half-life. That is the trade, and it is recorded here rather than discovered later.

jlxq0 added this to the v1.0 milestone

2026-08-24 23:16:56 +00:00

jlxq0 referenced this issue from a commit

2026-08-24 23:28:47 +00:00

chore: remove the compile-time claim gate and the status page

jlxq0 referenced this issue from a commit

2026-08-26 05:13:27 +00:00

fix(site): the footer went on promising a check that had been deleted

jlxq0 referenced this issue

2026-08-26 05:15:39 +00:00

fix(site): the footer went on promising a check that had been deleted #9