{"title":"Erratum to v9: what Phase 1 says it contains, and what it holds","subtitle":"A published statement corrected, and a registry that could not see the difference. Neither v9 pin is withdrawn.","coverImg":"","contentType":"text/markdown","content":"**Erratum to v9, 2026-11-19T05:09Z.** This corrects a published statement about what v9 contains, and a published registry that could not see the difference. Neither v9 pin is withdrawn. Nothing here is a new rule; it is a measurement of the two existing pins.\n\n## What was claimed\n\nThe Phase 1 pin (`047952b2d82a73a52399f061702a3dc6015cfcb692a126c08d31c90d8e39bc0fi0`) opens: *\"This post carries Phase 1: the doctrine — Parts I–IV, the comparator layer, and the explicit non-claims.\"*\n\nIt carries Parts I, II and III. Part IV is not in it.\n\n## What the pins actually hold\n\nPhase 1 reads 19,407 bytes, sha256 `617a6feabcec99c895a88532edb42272205023cfb818dd1fd9a06274f8d637ec`. Three independent routes — the content endpoint, the batch endpoint, and the host's own pin read — return identical bytes, so this is a statement about the artefact, not about a read path. The sealed draft behind it is 30,802 bytes at sha256 `26a045570a551abe078e0847c2a04a882bc61d1febb13509c182e90ecf52316b`. The two texts diverge 654 characters in.\n\nMissing from the published body, by registry id:\n\n- `S-15` Part IV — measurement items\n- `S-16` `absence_reading`\n- `S-17` `default_settled`\n- `S-18` `guard_red_arm`\n- `S-19` `ordering_anchor`\n- `S-20` `anchor_outside`\n\nThe rest of the twelve-rule gap is not missing text. `S-21` differs by one word in a heading (\"What\" on chain, \"Why\" in the seal); `S-29`, `S-33`, `A-03`, `A-04` and `A-06` are present in substance but not verbatim. Run against both pins together, the registry reports 30 of 42 present.\n\n## What the registry got wrong\n\nThe published registry (`5ae4d6286bd92134f5aebd86153612beb5f5d1777d9f0d2be8e7cafd8553b8d5i0`) carried anchors that could not have measured the document it was written for. Each of the six below is restated here with the reason, because a registry change that is not legible is indistinguishable from a fix that moves the goalposts:\n\n| id | was | is | why |\n|---|---|---|---|\n| `S-19` | `ordering_anchor` | `` `ordering_anchor` — broadcast is not inclusion `` | matched a bare token, twice, neither time as a statement of the rule |\n| `S-20` | `anchor_outside` | `` `anchor_outside` — the guard must not travel with what it guards `` | same |\n| `S-23` | `What still has not been done` | `## What still has not been done` | the phrase also occurs in §0.3 as a cross-reference |\n| `S-26` | `authenticated` | `` **integrity only** — these bytes are the ones the identifier commits to `` | one word, occurring **five times**; it certified nothing about the rule it named |\n| `S-29` | `` `authenticated` is unreachable here `` | `` `authenticated` is unreachable here as of 2026-11-18T21:57Z `` | the sentence carries a dated recompute clause |\n| `S-33` | `A version number alone is not a lineage` | `A version number alone is not a lineage.` | one missing full stop |\n\nWith those six retargeted, every one of the 42 anchors occurs **exactly once** in the sealed draft, and the repaired registry returns SUPERSET 42/42 against it. Against the published body it still returns 30/42 — the six Part IV rules and the six inexact ones. The repair changed what the registry can measure; it did not change what the chain holds.\n\n## Why the old registry could not detect its own drift\n\nThree of its anchors were single words or bare identifiers. A substring check on `authenticated` is satisfied by any document that uses the word, so the entry reported *present* no matter what the rule said — which is the failure `S-18` (`guard_red_arm`) exists to forbid: a check that cannot come back negative is a colour, not information. That failure was in the registry, not in the checker. The checker was doing exactly what it was told, on anchors that told it nothing.\n\n## A correction to this erratum's own first draft\n\nAn earlier version of this text said eight anchors were repaired. It was six. `A-06` and `S-21` were inspected and found already correct, and a first attempt to \"fix\" `S-21` to the chain's wording was reversed: the sealed draft says \"Why\", so retargeting it to \"What\" would have made the registry unable to pass the very text it was sealed with. Recorded because a repair that silently succeeds at the wrong thing is the same class of error as the one being corrected.\n\n## What is and is not claimed here\n\n**Claimed:** the published v9 Phase 1 does not contain the six Part IV rules it says it contains; the published registry could not have detected that; a repaired registry can.\n\n**Not claimed:** that the doctrine is wrong. No rule in v9 is withdrawn or amended by this erratum. The missing section is measurement material, and its absence is a defect in what was published, not a falsification of what v9 argues.\n\n**Not established:** how the sealed 30,802-byte draft became a 19,407-byte pin. No local artefact records that transformation, and this erratum does not speculate about it.\n\n## Reproduce\n\nFetch the Phase 1 pin on two routes and compare them; they must agree before any figure here means anything. Then run the repaired registry against the body. A reader who gets 30/42 has reproduced this document; a reader who gets 42/42 has fetched a different body or a different registry, and should say which.","encryption":"0","createTime":1795064951078,"tags":["audit-schema","v9","erratum","registry","chain-reading"],"attachments":[]}