Commit d44bd16
feat: declare that the registry serves time in abbreviated metadata
pnpr ships the per-version `time` map in its abbreviated packuments
precisely so pnpm's release-age check can read it there, and
`registrySupportsTimeField: true` is how a client is told to rely on
that. pnpm cannot assume it about an arbitrary registry, but the
benchmark can state it about its own.
Today the setting is measurably inert on the published rows: pnpm 12's
release-age verifier does not consult it - its timestamp ladder goes
abbreviated `modified` shortcut, local full-meta mirror, attestation
endpoint, full packument - so a package modified inside the cutoff
window still pays the full-metadata fallback (the lockfile row measured
4.6s with the flag as without; `trustLockfile: true` at 3.9s bounds
what the verifier could recover). The flag gates only time-based
resolution's up-front full fetch. Declared anyway because it is true,
and because the verifier learning to read per-version `time` from
abbreviated metadata (reported upstream) will then speed these rows up
with no benchmark change.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>1 parent 4dde06b commit d44bd16
1 file changed
Lines changed: 9 additions & 1 deletion
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
180 | 180 | | |
181 | 181 | | |
182 | 182 | | |
| 183 | + | |
| 184 | + | |
| 185 | + | |
| 186 | + | |
| 187 | + | |
| 188 | + | |
| 189 | + | |
| 190 | + | |
183 | 191 | | |
184 | 192 | | |
185 | | - | |
| 193 | + | |
186 | 194 | | |
187 | 195 | | |
188 | 196 | | |
| |||
0 commit comments