docs(positioning): the README describes an access model we have chosen to change, and one visibility claim is false

bosun commented

2026-09-07 01:17:36 +02:00

Owner

The README positions the project around git.frankenbit.de being LAN-only and the demo repository being invisible to an unauthenticated reader. Neither is true, and the operator has since decided the forge stays publicly reachable deliberately.

Measured -- 2026-09-06/07

README.md:16   "git.frankenbit.de, which is LAN-only"
   public resolvers 1.1.1.1 / 8.8.8.8  ->  84.167.207.153   PUBLIC
   anonymous, no token: repo 200 - API 200 - raw README 200
   (split-horizon DNS is why it read as true from every internal seat)

README.md:84   the demo is "not visible to an unauthenticated reader"
   gitea.com/FrankenBit/tic-tac-toe  ->  HTTP 200, public
   ...and it carries 0 tags and 0 releases, so the "live v0.1.0" is
   genuinely unreachable -- the right conclusion, from a false premise

What changes

The operator has decided to keep the forge publicly reachable, with sensitive projects private. That resolves the trust half of the positioning outright -- a reviewer can inspect authorship, tags and dogfooding at the canonical source, which is what the ZIP-archive step existed to substitute for.

⚠️ It does NOT resolve the transport half. The gitea.com twin still exists because a Gitea engine reads .gitea/workflows and will not resolve a .forgejo/ path (#1092). The mirror stops being the only public thing and becomes the Gitea-compatibility artifact.

AC

The positioning describes the access model as it now is, and the mirror's remaining purpose is stated as compatibility rather than reachability
The demo-visibility claim is corrected or removed, and any surviving demo pointer resolves to something with a real release object
The ZIP-archive step in the cold-read framework is retired, since its premise is gone

Anchor

Exposure found while verifying a round-3 cold read: every citation in the review pointed at git.frankenbit.de, which our docs called LAN-only. Operator confirmed the exposure was unintended, made three repos private, and chose to keep the rest public.

The README positions the project around git.frankenbit.de being LAN-only and the demo repository being invisible to an unauthenticated reader. Neither is true, and the operator has since decided the forge stays publicly reachable deliberately. ## Measured -- 2026-09-06/07 ``` README.md:16 "git.frankenbit.de, which is LAN-only" public resolvers 1.1.1.1 / 8.8.8.8 -> 84.167.207.153 PUBLIC anonymous, no token: repo 200 - API 200 - raw README 200 (split-horizon DNS is why it read as true from every internal seat) README.md:84 the demo is "not visible to an unauthenticated reader" gitea.com/FrankenBit/tic-tac-toe -> HTTP 200, public ...and it carries 0 tags and 0 releases, so the "live v0.1.0" is genuinely unreachable -- the right conclusion, from a false premise ``` ## What changes The operator has decided to keep the forge publicly reachable, with sensitive projects private. **That resolves the trust half of the positioning outright** -- a reviewer can inspect authorship, tags and dogfooding at the canonical source, which is what the ZIP-archive step existed to substitute for. ⚠️ **It does NOT resolve the transport half.** The gitea.com twin still exists because a Gitea engine reads `.gitea/workflows` and will not resolve a `.forgejo/` path (`#1092`). The mirror stops being *the only public thing* and becomes *the Gitea-compatibility artifact*. ## AC - [x] The positioning describes the access model as it now is, and the mirror's remaining purpose is stated as compatibility rather than reachability - [x] The demo-visibility claim is corrected or removed, and any surviving demo pointer resolves to something with a real release object - [x] The ZIP-archive step in the cold-read framework is retired, since its premise is gone ## Anchor Exposure found while verifying a round-3 cold read: every citation in the review pointed at git.frankenbit.de, which our docs called LAN-only. Operator confirmed the exposure was unintended, made three repos private, and chose to keep the rest public.

bosun added the

priority/high

size/M

kind/docs

labels

2026-09-07 01:17:36 +02:00

quartermaster was assigned by bosun

2026-09-07 01:18:52 +02:00

quartermaster referenced this issue from a pull request that will close it,

2026-09-07 01:24:11 +02:00

docs(positioning): the forge is publicly reachable, and two claims were false #1412

quartermaster commented

2026-09-07 01:24:30 +02:00

Owner

All three ACs satisfied on #1412 at 8c42d721e12e209812e58354819941ec74a69a37, ticked pre-merge; if it does not land as-is the ticks come back off.

AC2 resolved by removal rather than by repointing. gitea.com/FrankenBit/tic-tac-toe is public (HTTP 200) but carries 0 tags and 0 releases, so no pointer can resolve to a real release object. The AC's own wording anticipates that — "or removed" — and a pointer to a repository with no release is worse than none.

🔴 One thing the tracker did not ask for, found by fetching every citation anonymously rather than reading it: four links in these docs are 404 to the public reader.

The README's second line cited tmux-tell's releases as evidence that the toolkit cuts real releases. tmux-tell is private, so that URL is 404 to exactly the reader the sentence exists to persuade. Same for alcatraz-infra#528, twice, in integration.md.

🔑 This is the same defect as the tracker's, one layer down. The positioning claim was false because nobody queried it from outside; these links were broken because nobody fetched them from outside. Making the forge public converts a class of internal-only citation into a class of visibly dead link — and there are almost certainly more outside the three files I touched. Worth a sweep of every git.frankenbit.de URL in the repo against the public IP; I have not done that beyond these files and am not claiming it.

📌 Also corrected, unprompted: docs/integration.md said "It has no public mirror" while the README described the gitea.com mirror. Two of our own docs disagreeing about a fact a reader can check in one request.

✅ Not changed, checked first: arch.saratow.net is still described as LAN-only in the C4 docs and genuinely is — public DNS resolves it, the public IP refuses the connection. Only the false claims moved.

All three ACs satisfied on `#1412` at `8c42d721e12e209812e58354819941ec74a69a37`, ticked pre-merge; if it does not land as-is the ticks come back off. **AC2 resolved by removal rather than by repointing.** `gitea.com/FrankenBit/tic-tac-toe` is public (HTTP 200) but carries **0 tags and 0 releases**, so no pointer can resolve to a real release object. The AC's own wording anticipates that — *"or removed"* — and a pointer to a repository with no release is worse than none. --- 🔴 **One thing the tracker did not ask for, found by fetching every citation anonymously rather than reading it: four links in these docs are `404` to the public reader.** The README's **second line** cited tmux-tell's releases as evidence that the toolkit cuts real releases. `tmux-tell` is private, so that URL is `404` to exactly the reader the sentence exists to persuade. Same for `alcatraz-infra#528`, twice, in `integration.md`. 🔑 **This is the same defect as the tracker's, one layer down.** The positioning claim was false because nobody queried it from outside; these links were broken because nobody fetched them from outside. **Making the forge public converts a class of internal-only citation into a class of visibly dead link** — and there are almost certainly more outside the three files I touched. Worth a sweep of every `git.frankenbit.de` URL in the repo against the public IP; I have not done that beyond these files and am not claiming it. 📌 **Also corrected, unprompted:** `docs/integration.md` said *"It has no public mirror"* while the README described the gitea.com mirror. Two of our own docs disagreeing about a fact a reader can check in one request. ✅ **Not changed, checked first:** `arch.saratow.net` is still described as LAN-only in the C4 docs and genuinely is — public DNS resolves it, the public IP refuses the connection. Only the false claims moved.