tickets

5.4 KB

1. # Report 011 (w083): tickets#25 web part — short URLs /<slug> and /<slug>/<number>
3. **Where:** one commit on main in loreana:/media/STORAGE/projects/tickets.worldapi.org (on top of 0369106).
4. Not pushed, not deployed. No `.env` read. Ports used: 8760–8764 and 8767–8769 (8765/8766 not touched). Every server I
5. started was stopped by its PID. Nothing of mine is still running (checked with `ss -ltn`).
7. ## What changed
8. - **Pages:**
9. - a project is at `/<slug>`;
10. - a ticket is at `/<slug>/<number>`;
11. - every page link uses these. In the API rows, the `href` and `projectHref` values and the Markdown URL lines use them too.
12. - **301 redirects** to the short URL, using the project's CURRENT slug:
13. - `/projects/<slug>`;
14. - `/projects/<slug>/<n>`;
15. - `/projects/<slug>/tickets/<n>` (the form the conductor sent);
16. - `/tickets/<old n|uuid>`.
18. A URL that names nothing gets a 404 with no Location. That includes a CR/LF injection attempt.
19. - **Unchanged:**
20. - the API paths;
21. - `/projects/<slug>/settings`;
22. - the filter pages `/project/<slug>[/state/<s>]` and `/state/<s>`.
23. - **Reserved slugs** (`lib/projects.hl reservedSlugs`):
24. - **List:** every first path segment that a route answers (14 in project.hl, plus `__hl` and `components`), plus `my`, `me`, `assets`, `sign-in`, `sign-out`, `logout`, `settings`, `search`, `help`, `about`, `admin`, `robots.txt`, `.` and `..`.
25. - **Refused:** when a project is created and when a slug is changed (400 `field:"slug"`, "… is reserved").
26. - **Proposals:** a slug proposed from a title skips them (`Inbox` → `inbox-2`).
27. - **Gate check:** every route head in project.hl must be in the list.
28. - **Existing collisions: none.** The 11 live slugs are tickets.worldapi.org, ident.worldapi.org, gitoria.worldapi.org, antcolony, hybriel, layouts.worldapi.org, components.hybriel.worldapi.org, calendar.worldapi.org, notes.worldapi.org, termuplex and tracker.worldapi.org. No project has an old slug.
30. ## Route order: what hl:web really does
31. - The creator's rule holds: the FIRST matching route answers, in the order they are written (router.hl `match`).
32. - **Moving the wildcards to the bottom of OUR table is not enough.**
33. - `new WebFramework(…)` APPENDS hl:web's own URLs to the table: `/__hl/app.css`, `/__hl/emit`, `/__hl/sw.js`, the manifest, the runtime, and the component modules `/components/<file>.hl`.
34. - Even at the end of our table, `/:projectSlug/:ticketNumber` answered `/__hl/app.css` and `/components/ticket.hl` with a ticket page. I measured this.
35. - **Fix:** both slug routes stay at the end of the table in project.hl (marked `lastPage = true`). Right after `new WebFramework(…)`, project.hl moves them to the end of `server.router.routes`.
36. - **Why they stay in the table:** hl:web only serves a component's module when a route mounts it at construction. When I added the routes only afterwards, `/components/ticket.hl` returned HTML.
37. - **Dev watcher:** the order survives its re-analysis. I checked it with watch on and a changed component.
38. - **Control run:** without the move, the gate ends at 35 passed, 15 failed.
39. - **Recommendation:** a Hybriel issue saying that hl:web should match its own URLs before the app's routes. Then the move can go. I did not file it.
41. ## Tests
42. | test | result |
43. |---|---|
44. | gate `tests/browser.mjs` (`.scratch/w083/gate.sh`) | before: **249 passed, 0 failed**; after: **293 passed, 0 failed** (44 new checks: README "Test", "Mission 011 part") |
45. | `tests/connect.mjs` (temp copy, `.scratch/w083/connect.sh`) | **60 passed, 0 failed** (before and after) |
46. | live-data copy: new `tests/shorturls-realdata.mjs` | **187 passed, 0 failed** (see below) |
47. | the ticket's own test, by curl on the copy | `/tracker.worldapi.org` and `/tracker.worldapi.org/31` → 200 (titles right); both old forms → 301 to them |
48. | API, old vs new code on the same copy (`.scratch/w083/apicompare.py`) | **0 different** (see below) |
49. | `tests/letcount.py` | never-reassigned `let` = 0 |
51. **Live-data copy:** taken from Byrodin on 2026-10-03 around 20:20 and deleted after the runs.
52. - **Projects:** all 11. Each `/<slug>` shows the heading with its title, and `/projects/<slug>` returns 301 to it.
53. - **Tickets:** 20, round robin over the projects. For each:
54. - the short URL shows the ticket page with its subject;
55. - `/projects/<s>/<n>`, `/projects/<s>/tickets/<n>` and `/tickets/<old n>` all return 301 to it, and following the redirect gives 200;
56. - the Markdown URL line is the short URL.
58. **API comparison:**
59. - 1,348 answers (JSON and Markdown).
60. - 17 are identical.
61. - 1,331 differ ONLY in the page links (the `href`/`projectHref` values, the Markdown URL lines). `apiHref` and `settingsHref` are unchanged.
63. ## Found
64. - **hl:web route order:** see above. Written into STATUS "Lessons".
65. - **My own slip, caught:** a `setsid … &` inside a `( … )` subshell forks, so `$!` was not the server's PID.
66. - A leftover test server kept port 8763, and my second comparison run used it.
67. - I found it, checked by its cwd and environment that it was mine, killed it by PID, and re-ran the comparison.
68. - This is written into STATUS "Lessons".
70. ## Not done / open
71. - **The API part** (`/api/tickets/<x>`, `/api/projects/<id>`), as the mission said. The creator still has to say whether `<x>` is a global id.
72. - **Not deployed:** deploying is the architect's job (`./deploy.sh`). There is no data change.
73. - **Not touched:**
74. - DECISIONS.md: it belongs to the librarian. Its links use the old form and still work through the 301.
75. - The old one-off tests `tests/browser-pre20-parked.txt` and `tests/short-id-switch.mjs`.