|
| 1 | +# Authorizations |
| 2 | + |
| 3 | +This file records operator authorizations to edit corpus text. It exists because an authorization |
| 4 | +that lives only in a chat turn, a commit message, or a signature row is not discoverable later by |
| 5 | +anyone reading the corpus. |
| 6 | + |
| 7 | +**What this file is not.** It is not a section or appendix of any contract or ruling. It is not the |
| 8 | +gap register (`open_contract_gaps.md`), which tracks what a contract fails to enforce, and it is not |
| 9 | +the corrections register (`corrections.md`), which records defects in text and — by its own standing |
| 10 | +rule at `corrections.md:12-13` — **cannot** authorize an edit. An authorization is a third act: it |
| 11 | +grants permission to change text, names exactly what may change, and closes when the authorized |
| 12 | +commit lands. |
| 13 | + |
| 14 | +**Standing rule.** An entry here authorizes only what it names. Nothing in this file is a signature, |
| 15 | +and no entry ticks a sign-off box. An authorization that has been exercised remains on the record; |
| 16 | +it is not deleted once used. |
| 17 | + |
| 18 | +**Why this file was created on 2026-08-23.** The two prior pre-signature amendments to Contract 001 |
| 19 | +were authorized nowhere in the corpus. `425aa09` (`:114`, `:153`) carries its authority only in its |
| 20 | +commit message — *"Operator ruling: match :114 to :128's count basis rather than tie it to a support |
| 21 | +threshold."* `0c83577` (`:57`) carries no authorization line at all, and is known to have been |
| 22 | +authorized only because §6's signature (`contract_001_domain_resolution.md:271`) later cites it as |
| 23 | +*"the standing precedent that the route is available in this corpus."* Grepping `docs/` for either |
| 24 | +hash returns two signature rows and nothing else. Those two amendments are not reopened by this |
| 25 | +file; the gap in the record is noted, not retroactively filled. |
| 26 | + |
| 27 | +--- |
| 28 | + |
| 29 | +## AUTH-001 |
| 30 | + |
| 31 | +``` |
| 32 | +Authorizes: editing Contract 001 §7, and only §7 |
| 33 | +Target lines: :245, :246, :248 — three edits, enumerated below |
| 34 | +Route: amend-then-sign (route A), per the precedent at 0c83577 |
| 35 | +Granted by: Kevin Brown, operator |
| 36 | +Granted at: 2026-08-23, in session |
| 37 | +Basis: CR-002 findings 1, 2, 3 and CR-003, all recorded in |
| 38 | + corrections.md; §7 is UNSIGNED, so no attestation is |
| 39 | + disturbed by an edit made before signature |
| 40 | +Status: granted; NOT YET EXERCISED — no edit has been made to |
| 41 | + contract_001_domain_resolution.md as of this recording |
| 42 | +``` |
| 43 | + |
| 44 | +**Why route A and not route C.** Route C — relocating §7 out of the contract entirely, so that |
| 45 | +§1's self-test (`contract_001_domain_resolution.md:41-42`) no longer reaches it — was presented and |
| 46 | +not taken. Route A keeps `:239`'s pattern where a reader of the contract meets it, and §6's own |
| 47 | +stated test for when amendment is available is met: these are local text repairs, not the |
| 48 | +*"substantive invariant design with implementation consequences"* that §6's signature |
| 49 | +(`contract_001_domain_resolution.md:271`) gives as its reason for declining the route for GAP-4 and |
| 50 | +GAP-5. No invariant is touched by any edit below. |
| 51 | + |
| 52 | +### The three authorized edits |
| 53 | + |
| 54 | +**Edit 1 — `:245`, row 2's "what is preserved" cell.** Discharges CR-002 finding 1. |
| 55 | + |
| 56 | +- From: `the meaning of a category` |
| 57 | +- To: `the integrity of what members inherit from a family` |
| 58 | + |
| 59 | +Grounded in `ruling_002_family_taxonomy_integrity.md:96` and the mechanism amendment at |
| 60 | +`ruling_002_family_taxonomy_integrity.md:101-108`. Deliberately neutral between the two inheritance |
| 61 | +channels: the amendment warns at `ruling_002_family_taxonomy_integrity.md:107-108` that a future |
| 62 | +reader must not read the grounding sentence as current behavior, so the cell may not say |
| 63 | +"grounding"; "classification" alone would drop the prospective half the amendment states is a |
| 64 | +reason for the same rule. |
| 65 | + |
| 66 | +**Edit 2 — `:246`, row 3's rule cell.** Discharges CR-002 finding 2. |
| 67 | + |
| 68 | +- From: `ambiguity is preserved, never collapsed` |
| 69 | +- To: `ambiguity is preserved, never collapsed; selecting arbitrarily is prohibited` |
| 70 | + |
| 71 | +I3 (`contract_001_domain_resolution.md:221-222`) closes *"Selecting arbitrarily is prohibited"* — |
| 72 | +the clause §6's signature (`:271`) identifies as operative and without which it holds I3 is not |
| 73 | +adequately represented. Row 3's "what is preserved" cell was not flagged by Beat 1 and is **not** |
| 74 | +authorized for change. |
| 75 | + |
| 76 | +**Edit 3 — `:248`, the synthesis sentence.** Discharges CR-002 finding 3 **and** CR-003 in a single |
| 77 | +amendment. |
| 78 | + |
| 79 | +- From: `In each case the system is permitted to be *incomplete* but never permitted to be *falsely certain*.` |
| 80 | +- To: `Each of the three rules accepts a less resolved result rather than a wrongly resolved one: a duplicate left unmerged, a family left unadmitted, a domain left unselected.` |
| 81 | + |
| 82 | +**One edit, not two, and the reason is not merely that both defects share a sentence.** A bare |
| 83 | +strike of *"in each case"* would repair jurisdiction — the distribution onto row 2 disappears and |
| 84 | +the subject reverts to I6's shape — but would leave *"the system is permitted to be incomplete"* |
| 85 | +asserted without restriction and without any cited support, which is a worse sourcing posture than |
| 86 | +CR-002 finding 3 describes rather than a repair of it. Each defect's fix redefines the other's |
| 87 | +target, so the two cannot be sequenced in either order and must be one amendment justified by both |
| 88 | +findings. |
| 89 | + |
| 90 | +The replacement was checked against four things: |
| 91 | + |
| 92 | +1. **Jurisdiction (CR-003).** The subject is *the three rules*, not *the system*, and no deontic |
| 93 | + verb remains. The sentence reports rather than conditions, placing it in the same observational |
| 94 | + register as `:245`, which already survives `:41-42`. |
| 95 | +2. **`:239`'s stated purpose.** The section's value is the pattern. The replacement names all three |
| 96 | + rows in table order, so the sentence stays tied to the table it summarizes. |
| 97 | +3. **CR-002 finding 3.** The original's error was choosing *incomplete* as the unifying term, which |
| 98 | + fits rows 2-3 but not row 1, where Ruling 001 §1 licenses noise — over-completeness |
| 99 | + (`ruling_001_canonicalization_policy.md:31`). *Less resolved* covers all three: an unmerged |
| 100 | + duplicate is an identity question the system declined to decide. Row 1 is no longer an exception. |
| 101 | +4. **What Beat 1 found sound is preserved.** *"Never falsely certain"* held across all three rows; |
| 102 | + *"rather than a wrongly resolved one"* carries it forward. |
| 103 | + |
| 104 | +### This authorization does NOT authorize |
| 105 | + |
| 106 | +- Any edit to `contract_001_domain_resolution.md` outside `:245`, `:246`, and `:248`. |
| 107 | +- Any wording at those three lines other than the wording set out above. A different repair requires |
| 108 | + a different authorization. |
| 109 | +- Ticking §7's sign-off box (`:272`), or any other box. Amendment and signature are separate acts, |
| 110 | + and §7 remains unsigned when the authorized edits land. |
| 111 | +- Any edit to `:272`'s existing text, whose rejection as clearance is recorded in CR-003. |
| 112 | +- Reopening, amending, or correcting CR-002 or CR-003, or any signed row. |
| 113 | +- Any edit to Ruling 001 or Ruling 002, both LOCKED. |
| 114 | +- Route C. Relocating §7 out of the contract is not authorized by this entry. |
| 115 | + |
| 116 | +### Verification required before the authorized commit lands |
| 117 | + |
| 118 | +- Every line of `contract_001_domain_resolution.md` other than `:245`, `:246`, and `:248` byte-identical |
| 119 | + to the pre-edit file, so no signature's anchors shift — the check `0c83577`'s commit message |
| 120 | + records ("zero anchor shift"). |
| 121 | +- File length unchanged at 275 lines. |
| 122 | +- Every `:NNN` reference in the amended §7 re-resolved by extraction against the post-edit file. |
| 123 | + |
| 124 | +### Known consequence, recorded not addressed |
| 125 | + |
| 126 | +CR-002 and CR-003 quote `:245`, `:246`, and `:248` as they read **before** these edits. Those |
| 127 | +quotations become historical on the authorized commit. Both entries are true as of their recording |
| 128 | +dates and are left as recorded, per the standing disposition this corpus applies to signatures. |
| 129 | +Correcting them would be a separate act, and this authorization does not permit it. |
| 130 | + |
| 131 | +**Acknowledgment** — an acknowledgment, not a section signature. |
| 132 | + |
| 133 | +``` |
| 134 | +Acknowledged (operator): Kevin Brown |
| 135 | +Acknowledged at: 2026-08-23 19:20 EDT |
| 136 | +``` |
| 137 | + |
| 138 | +Statement: *"AUTH-001 is acknowledged. Creating a third instrument rather than forcing the |
| 139 | +authorization into `corrections.md` is correct, and it was checked into existence rather than |
| 140 | +argued into it: grepping for the two prior amendment hashes and finding nothing outside a commit |
| 141 | +message and two signature rows demonstrates that authorization living only in ephemeral text is |
| 142 | +not a hypothetical risk — it has already happened twice. Recording that `425aa09` and `0c83577` |
| 143 | +lack discoverable authorization, without retroactively supplying one, is right; the record should |
| 144 | +show the gap existed rather than paper over it now that the fix is available. The authorization is |
| 145 | +bounded to wording and not only to line numbers, and the does-not-authorize list is part of what is |
| 146 | +granted, not commentary on it. **AUTH-001 pre-clears the permission question only. It does not |
| 147 | +pre-clear the wording.** The §7 amendment commit remains subject to diff review of its proposed |
| 148 | +replacement text before it lands; permission and wording-is-good are separate reviews, even though |
| 149 | +this entry lists specimens."* |
0 commit comments