Commit 090dcbc
authored
Most of the shelf no longer installs — measured, charted, and tripwired (#5)
* site: most of the shelf no longer installs, measured
This site's promise is the install decision, and until today it only ever
asked whether a plugin declares an install path. It never asked whether that
path still resolves. It usually does not.
## The measurement
`scripts/measure-installability.mjs` reads every npm name in the registry off
registry.npmjs.org and tests each `@deepseek-ai/dsh*` range against every
published dsh version. No GitHub budget, no clone.
Of 1,280 published packages that name a dsh version:
- 79 (6%) accept dsh 0.1.2-rc.1, what `npx @deepseek-ai/dsh` gives you
- 818 accept nothing newer than the 0.1.0 line
- 182 accept no published dsh at all
- 138 use a wildcard: never breaks, never protects
## Why
npm semver never lets a prerelease satisfy a caret with a different version
tuple, and dsh has only ever shipped prereleases. `^0.1.1-rc.1` matches
0.1.1-rc.2 and stops dead; `^0.1.2-alpha.1` matches 0.1.2-rc.1, because the
tuple is the same. The rule is the tuple, not the caret, and it is invisible
in a diff.
The loud failure is ERESOLVE and you would fix that. The quiet one is what
reaches people: install a stale plugin alone and npm hoists the old
`@deepseek-ai/dsh-*` copies to the root, pushing the harness's own into nested
node_modules. The install succeeds and the plugin imports a different harness
than the host runs. Measured on one package: 13 harness packages at two
versions at once, no warning.
Three of our own four plugins were in that column this morning. The section
says so.
## Two tripwires, one of which I got wrong first
- `check-claims` now fails when the installability census names a different
dsh than the ecosystem census does. A census stale by version is not stale
by date, and the 14-day rule cannot see it. **It caught one immediately:**
data/ecosystem.json still said `npx` gives you 0.1.1-rc.2, three days after
0.1.2-rc.1 took the `latest` tag. Both censuses re-measured.
- The first version of that check sat BELOW the `if (fail.length)` gate, so it
collected failures nothing ever read — green in both directions. Moved above
the gate and mutation-checked: it fails on a wrong version now.
- data/installability.json is allowlisted in .gitignore and re-measured in the
deploy job. A census the deploy does not re-run is the failure this repo
already had once, for twelve days.
* data: rebuild against the merged registries (13,381 plugins, 486 themes)
941 entries now say they carry no proof, up from 927: the plugins sweep
demoted 84 rows whose only receipt named cordis rather than the harness.1 parent 4e46598 commit 090dcbc
13,393 files changed
Lines changed: 13800 additions & 12098 deletions
File tree
- .github/workflows
- data
- public/_data
- p
Some content is hidden
Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
89 | 89 | | |
90 | 90 | | |
91 | 91 | | |
| 92 | + | |
| 93 | + | |
| 94 | + | |
| 95 | + | |
| 96 | + | |
92 | 97 | | |
93 | 98 | | |
94 | 99 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
19 | 19 | | |
20 | 20 | | |
21 | 21 | | |
| 22 | + | |
| 23 | + | |
| 24 | + | |
22 | 25 | | |
23 | 26 | | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
1 | 1 | | |
2 | 2 | | |
3 | | - | |
| 3 | + | |
4 | 4 | | |
5 | 5 | | |
6 | 6 | | |
7 | | - | |
8 | | - | |
| 7 | + | |
| 8 | + | |
9 | 9 | | |
10 | | - | |
11 | | - | |
12 | | - | |
13 | | - | |
14 | | - | |
15 | | - | |
16 | | - | |
17 | | - | |
18 | | - | |
| 10 | + | |
| 11 | + | |
19 | 12 | | |
20 | | - | |
21 | | - | |
| 13 | + | |
| 14 | + | |
22 | 15 | | |
23 | 16 | | |
24 | 17 | | |
| |||
42 | 35 | | |
43 | 36 | | |
44 | 37 | | |
45 | | - | |
| 38 | + | |
46 | 39 | | |
47 | 40 | | |
48 | 41 | | |
| |||
54 | 47 | | |
55 | 48 | | |
56 | 49 | | |
57 | | - | |
| 50 | + | |
58 | 51 | | |
59 | 52 | | |
60 | 53 | | |
61 | | - | |
| 54 | + | |
62 | 55 | | |
63 | 56 | | |
64 | 57 | | |
65 | | - | |
| 58 | + | |
66 | 59 | | |
67 | 60 | | |
68 | 61 | | |
69 | | - | |
| 62 | + | |
70 | 63 | | |
71 | 64 | | |
72 | 65 | | |
73 | 66 | | |
74 | 67 | | |
75 | 68 | | |
76 | 69 | | |
77 | | - | |
| 70 | + | |
78 | 71 | | |
79 | 72 | | |
80 | 73 | | |
81 | 74 | | |
82 | 75 | | |
83 | 76 | | |
84 | 77 | | |
85 | | - | |
| 78 | + | |
86 | 79 | | |
87 | 80 | | |
88 | 81 | | |
89 | 82 | | |
90 | 83 | | |
91 | 84 | | |
92 | 85 | | |
93 | | - | |
| 86 | + | |
94 | 87 | | |
95 | 88 | | |
96 | 89 | | |
97 | | - | |
| 90 | + | |
98 | 91 | | |
99 | 92 | | |
100 | 93 | | |
101 | | - | |
| 94 | + | |
102 | 95 | | |
103 | 96 | | |
104 | 97 | | |
105 | | - | |
| 98 | + | |
106 | 99 | | |
107 | 100 | | |
108 | 101 | | |
109 | | - | |
| 102 | + | |
110 | 103 | | |
111 | 104 | | |
112 | 105 | | |
113 | | - | |
| 106 | + | |
114 | 107 | | |
115 | 108 | | |
116 | 109 | | |
117 | | - | |
| 110 | + | |
118 | 111 | | |
119 | 112 | | |
120 | 113 | | |
121 | | - | |
| 114 | + | |
122 | 115 | | |
123 | 116 | | |
124 | 117 | | |
125 | | - | |
| 118 | + | |
126 | 119 | | |
127 | 120 | | |
128 | 121 | | |
129 | | - | |
| 122 | + | |
| 123 | + | |
| 124 | + | |
| 125 | + | |
| 126 | + | |
| 127 | + | |
| 128 | + | |
| 129 | + | |
| 130 | + | |
| 131 | + | |
| 132 | + | |
| 133 | + | |
| 134 | + | |
130 | 135 | | |
131 | 136 | | |
132 | 137 | | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| 2 | + | |
| 3 | + | |
| 4 | + | |
| 5 | + | |
| 6 | + | |
| 7 | + | |
| 8 | + | |
| 9 | + | |
| 10 | + | |
| 11 | + | |
| 12 | + | |
| 13 | + | |
| 14 | + | |
| 15 | + | |
| 16 | + | |
| 17 | + | |
| 18 | + | |
| 19 | + | |
| 20 | + | |
| 21 | + | |
| 22 | + | |
| 23 | + | |
| 24 | + | |
| 25 | + | |
| 26 | + | |
| 27 | + | |
| 28 | + | |
| 29 | + | |
0 commit comments