This page defines the status words PICurv documentation uses to describe a capability. The words are load-bearing: they tell a reader whether a feature can be relied on for published work, and they gate what documentation a feature owes before it may claim the next level.
Use these words exactly. Do not invent synonyms, and do not leave a public capability unlabelled.
make audit-capability fails on a missing status, a status outside this list, a non-selectable status on a reachable value, or a documentation part the claimed status owes. The same vocabulary applies to subsystems, where make audit-subsystems enforces the obligations each status carries (Documentation Extension Framework ยง4).Status describes what the project claims about a capability today.
| Status | Meaning | May appear in user reference? |
|---|---|---|
planned | Designed or proposed; no usable implementation exists. | No - design/specification pages only |
internal | Implemented, but not exposed as a user-selectable option. | No - developer documentation only |
experimental | User-selectable, but not established for production or published work. | Yes, with explicit limitations |
supported | User-selectable, documented, evidenced, and suitable for production use. | Yes |
deprecated | Still functional, superseded; scheduled for removal. | Yes, with migration guidance |
removed | No longer present; retained only as history and migration guidance. | Changelog/history only |
known-defective | Reachable by a user but known to produce incorrect results. | Yes - disclosure is mandatory |
known-defective is the one status with a hard placement requirement:
A capability may be known-defective while the runtime still accepts it. The documentation states the defect; whether the runtime should refuse the option is a separate engineering decision and is tracked outside the documentation program.
Status and evidence are independent axes and must not be collapsed into one ladder. Production use does not imply analytical verification, and a validation comparison does not imply restart equivalence.
Evidence is recorded as facets, each linking to the test, example, report, or dataset that establishes it:
A capability with no evidence facet is implemented only. That is a legitimate state to record; it is not a legitimate state to hide.
An observation measured against code that has since changed is neither current behavior nor a defect. Record it as:
State when it was measured and what changed underneath it. Do not silently delete such an observation, and do not let it keep circulating as current fact. When it is rerun, replace it with the new measured result.