Compare commits

...
Author SHA1 Message Date
mesh-tools f0a940463b Merge pull request 'Design 47 G: the lead's answers A-463a and A-464d (from the review)' (#658) from design/47-G-lead-answers-72 into main 2026-11-17 01:08:47 +00:00
jschoubben 5385c798c0 Answer A-463a and A-464d: a pending merge is not behind, and every source move is forward 2026-11-17 02:08:39 +01:00
mesh-tools cf08f6e69b Merge pull request 'Design 47 G: the lead's answer A-464c (the gate's source moves a head only forward)' (#657) from design/47-G-lead-answers-71 into main 2026-11-17 01:01:11 +00:00
jschoubben 163fe51f96 Answer A-464c: the gate's source moves a head only forward, and only when it changes 2026-11-17 02:00:59 +01:00
mesh-tools c81a186a3d Merge pull request 'Design 47 G: the lead's answer A-464b (a merge the gate did not make)' (#656) from design/47-G-lead-answers-70 into main 2026-11-17 00:57:23 +00:00
jschoubben acd7f1f588 Answer A-464b: what makes a merge one the gate did not make, and where that is said 2026-11-17 01:57:14 +01:00
mesh-tools 6664193aac Merge pull request 'Design 47 G: the lead's answer A-464a (a read tool for the indexer's takeins state)' (#655) from design/47-G-lead-answers-69 into main 2026-11-17 00:50:39 +00:00
jschoubben d47f617ad5 Answer A-464a: a read tool for the indexer's takeins state, so a hand take-in is seen 2026-11-17 01:50:25 +01:00
mesh-tools 84ef09d4b4 Merge pull request 'Design 47 G: the lead's answer A-464 (the gate's source follows every merge)' (#654) from design/47-G-lead-answers-68 into main 2026-11-17 00:43:10 +00:00
jschoubben 709ad9de59 A-464: name mesh-host in a code span, as the words test asks 2026-11-17 01:42:59 +01:00
jschoubben a4d7d44cd0 Answer A-464: the gate's source follows every merge into a watched branch 2026-11-17 01:42:43 +01:00
mesh-tools 47332934b8 Merge pull request 'Design 47 G: the lead's answer A-463 (the indexer takes in a merge it never took in)' (#653) from design/47-G-lead-answers-67 into main 2026-11-17 00:40:25 +00:00
jschoubben b6c99f171d Answer A-463: the indexer takes in a merge it never took in, and says so 2026-11-17 01:40:16 +01:00
mesh-tools e34c74c2f8 Merge pull request 'Design 47 G: A-459b withdraws A-459 and A-459a; A-462b' (#652) from design/47-G-lead-answers-66 into main 2026-11-17 00:32:53 +00:00
jschoubben f44b0fc90b Answer A-459b (withdraw A-459, A-459a: rule 7 reads the node's own composition) and A-462b 2026-11-17 01:32:44 +01:00
mesh-tools c91aafb6dd Merge pull request 'Design 47 G: the lead's answer A-457b (republish-max-versions goes)' (#651) from design/47-G-lead-answers-65 into main 2026-11-17 00:30:28 +00:00
jschoubben 1209008aba Answer A-457b: the builder's republish-max-versions setting goes with the path it bounded 2026-11-17 01:30:16 +01:00
mesh-tools 25449d1d4f Merge pull request 'Design 47 G: the lead's answer A-456b (the release record is bounded; a publish item ends with its release)' (#650) from design/47-G-lead-answers-64 into main 2026-11-17 00:24:02 +00:00
jschoubben dcaad0c6a4 Answer A-456b: the release record is bounded per module; a publish item ends with its release or its reservation 2026-11-17 01:23:38 +01:00
mesh-tools 66b3d0c7f6 Merge pull request 'Design 47 G: correct A-459a (composed is followed through the declared read)' (#649) from design/47-G-lead-answers-63 into main 2026-11-17 00:22:12 +00:00
jschoubben 92929d55ef Correct A-459a: rule 7 follows composed through the read the assigner's manifest declares 2026-11-17 01:22:00 +01:00
mesh-tools 0bb30e6bc9 Merge pull request 'Design 47 G: the lead's answer A-461b (an absent placed build is the trunk build)' (#648) from design/47-G-lead-answers-62 into main 2026-11-17 00:18:06 +00:00
jschoubben e69d2693de Answer A-461b: an absent placed build is the trunk build from mesh-catalog's module verb 2026-11-17 01:17:57 +01:00
mesh-tools b90489afc7 Merge pull request 'Design 47 G: the lead's answer A-460a (ace's refusal was a hand act; no retry added)' (#647) from design/47-G-lead-answers-61 into main 2026-11-17 00:11:52 +00:00
jschoubben bca45ea0df Answer A-460a: ace's refusal was a hand act; a failed install is tried again when its want changes 2026-11-17 01:11:39 +01:00
mesh-tools f48d5f5091 Merge pull request 'Design 47 G: the lead's answers A-456a, A-457a, A-461a' (#646) from design/47-G-lead-answers-60 into main 2026-11-17 00:11:21 +00:00
jschoubben 2ff96d1b7e Answer A-461a: another node's module is laid from the assignment's placed build 2026-11-17 01:11:09 +01:00
jschoubben e68d3a892e Answer A-456a and A-457a: a released build counts as reserved; the publish retries on the builder's own re-ask state 2026-11-17 01:10:59 +01:00
mesh-tools de9d3378af Merge pull request 'Design 47 G: the lead's answer A-462a (no entry in the followed catalog; the assigner asks the indexer nothing)' (#645) from design/47-G-lead-answers-59 into main 2026-11-17 00:10:20 +00:00
jschoubben 547006dd59 Answer A-462a: no entry means none in the followed catalog; the assigner asks the indexer nothing 2026-11-17 01:10:07 +01:00
mesh-tools 2945e57a4a Merge pull request 'Design 47 G: the lead's answer A-459a (rule 7 hears composed per module and fingerprint)' (#644) from design/47-G-lead-answers-58 into main 2026-11-17 00:09:49 +00:00
jschoubben fc5ec32b90 Answer A-459a: rule 7 hears composed per module and fingerprint from the node-engine's events 2026-11-17 01:09:38 +01:00
mesh-tools 242bae0d0c Merge pull request 'Design 47 G: the lead's answers A-454a, A-456..A-462 (the delivery path after a passed pipeline)' (#643) from design/47-G-lead-answers-57 into main 2026-11-17 00:05:31 +00:00
jschoubben 928aa7e5ba Answer A-454a, A-456..A-462: close the delivery path's six gaps after a passed pipeline 2026-11-17 01:05:14 +01:00
mesh-tools 4d4815c83d Merge pull request 'Design 47 G: the lead's answer A-455 (a stale reservation is taken over; builder_reserve_done)' (#642) from design/47-G-lead-answers-56 into main 2026-11-16 23:50:58 +00:00
jschoubben 975f83f045 Answer A-455: a newer trunk build takes over a stale reservation; builder_reserve_done for a done build 2026-11-17 00:50:50 +01:00
mesh-tools acc54c16b9 Merge pull request 'Design 47 G: correct A-453a (a held have has not moved past its hold)' (#641) from design/47-G-lead-answers-55 into main 2026-11-16 23:49:01 +00:00
jschoubben 77cf51b4c7 Correct A-453a: 'at another fingerprint' applies only to a have that is not held 2026-11-17 00:48:47 +01:00
mesh-tools 5f90d599c6 Merge pull request 'Design 47 G: the lead's answer A-454 (an upstream image is no part of the want)' (#640) from design/47-G-lead-answers-54 into main 2026-11-16 23:47:36 +00:00
jschoubben 987283dd30 Answer A-454: an upstream image a manifest pins is no part of the want 2026-11-17 00:47:28 +01:00
mesh-tools 0da67d6323 Merge pull request 'Design 47 G: the lead's answer A-453a (a have mid-move is not 'no longer held')' (#639) from design/47-G-lead-answers-53 into main 2026-11-16 23:46:13 +00:00
jschoubben 2ecb831355 Answer A-453a: a have mid-move is not 'no longer held' 2026-11-17 00:46:04 +01:00
mesh-tools 59d9f397e2 Merge pull request 'Design 47 G: the lead's answer A-453 (a backup ask only for a failed backup)' (#638) from design/47-G-lead-answers-52 into main 2026-11-16 23:37:41 +00:00
jschoubben 7b1607d933 Answer A-453: the operator is asked about a backup only when a backup failed 2026-11-17 00:37:32 +01:00
mesh-tools bfc7272b3c Merge pull request 'Design 47 G after delivery: retire held/kept for a module left on its build; hollow composition pass' (#637) from design/47-G-after-words into main 2026-11-16 23:17:48 +00:00
jschoubben 37fdc292f7 After G: retire held and kept for a module left on its build; a composition that judged nothing is no pass 2026-11-17 00:17:38 +01:00
mesh-tools 21ab7b1bdf Merge pull request 'Design 47 G: the lead's answer A-452 (an unanswered requirement costs its consumer, not the machine)' (#636) from design/47-G-lead-answers-51 into main 2026-11-16 23:15:03 +00:00
jschoubben e17cffa59e Answer A-452: an unanswered requirement costs its consumer, not the machine 2026-11-17 00:14:46 +01:00
mesh-tools dcdfbc12a6 Merge pull request 'Design 47 G: the lead's answer A-451 (composition with a node-engine not heard running)' (#635) from design/47-G-lead-answers-50 into main 2026-11-16 13:53:44 +00:00
jschoubben 7dd6563ab5 Answer A-451: a machine whose node-engine is not heard running composes with the revision's node-engine entry 2026-11-16 14:53:14 +01:00
mesh-tools eb5cc23a86 Merge pull request 'Design 47 G: the lead's answer A-237c (inline input stays for the checker until G)' (#634) from design/47-G-lead-answers-49 into main 2026-11-16 13:47:23 +00:00
jschoubben d07de2840d Answer A-237c: the token-less inline input stays for the checker until G removes it 2026-11-16 14:47:11 +01:00
mesh-tools 2ab0507440 Merge pull request 'Design 47 G: the lead's answer A-237b (compose-candidate sends its input in chunks)' (#633) from design/47-G-lead-answers-48 into main 2026-11-16 13:35:32 +00:00
jschoubben c47f9ea48e Answer A-237b: compose-candidate sends its input in chunks 2026-11-16 14:35:22 +01:00
mesh-tools dd462aa1ff Merge pull request 'Design 47 G: the lead's answer A-450 (the node-engine's account in the catalog's caller check)' (#632) from design/47-G-lead-answers-47 into main 2026-11-16 13:12:41 +00:00
jschoubben c0736710e8 Answer A-450: the node-engine's account counts as the module holding node-engine for the catalog's caller check 2026-11-16 14:12:28 +01:00
mesh-tools bedc92e251 Merge pull request 'Design 47 G: the lead's answer A-226c (a run's key reads candidates.run-<id>)' (#631) from design/47-G-lead-answers-46 into main 2026-11-16 12:59:58 +00:00
jschoubben d93dddf5b3 Answer A-226c: a run's key reads candidates.run-<id> 2026-11-16 13:59:47 +01:00
mesh-tools 2a52d91cc6 Merge pull request 'Design 47 G: the lead's answer A-226b (compose-candidate reads a revision key and a run key)' (#630) from design/47-G-lead-answers-45 into main 2026-11-16 12:55:03 +00:00
jschoubben ecaf070846 Answer A-226b: compose-candidate reads a branch revision key and a run key too 2026-11-16 13:54:52 +01:00
mesh-tools 2631a11447 Merge pull request 'Principles 11 and 12, and the reviewers ask them' (#629) from principles/configurable-and-own-files into main 2026-11-16 12:53:40 +00:00
jschoubben 39f7ae7203 Reviewers: a hard-coded value and a file written for another module are findings (principles 11 and 12) 2026-11-16 13:53:30 +01:00
jschoubben 6517bbebf0 Principles 11 and 12: configurable, never hard-coded; a module writes only its own 2026-11-16 13:51:28 +01:00
mesh-tools 4de493a784 Merge pull request 'Design 47 G: the lead's answer A-449a (the indexer pins a take-in's revision for its pipelines)' (#628) from design/47-G-lead-answers-44 into main 2026-11-16 12:44:33 +00:00
jschoubben 79214618f3 Answer A-449a: the indexer pins a take-in's revision for its pipelines when it keeps it 2026-11-16 13:44:23 +01:00
mesh-tools 8fa037074c Merge pull request 'Design 47 G: the lead's answers A-448 and A-449 (a take-in names its revision; a pipeline pins it)' (#627) from design/47-G-lead-answers-43 into main 2026-11-16 12:33:23 +00:00
jschoubben f07c337c22 Answer A-448 and A-449: a take-in names its revision; a pipeline pins it 2026-11-16 13:33:12 +01:00
mesh-tools d595d08042 Merge pull request 'After G: tests against the composed bus, kept events heard or loud, one way of asking work' (#626) from design/47-G-after-list-2 into main 2026-11-16 12:01:14 +00:00
jschoubben eaa6de7faf After G: tests against the composed bus, kept events heard or loud, one way of asking work 2026-11-16 13:01:03 +01:00
mesh-tools 2d4b34c51c Merge pull request 'Design 47 G: the lead's answer A-447e (the checker is removed at G, not rebuilt per module)' (#625) from design/47-G-lead-answers-42 into main 2026-11-16 10:39:32 +00:00
jschoubben d4c31d4ff0 Answer A-447e: the checker is removed at G, not rebuilt per module; the no-dir refusal lands with its removal 2026-11-16 11:39:23 +01:00
mesh-tools 71cb172953 Merge pull request 'Design 47 G: the lead's answer A-447d (the checker fetches per changed module)' (#624) from design/47-G-lead-answers-41 into main 2026-11-16 10:38:20 +00:00
jschoubben fa74c6d594 Answer A-447d: the checker fetches per changed module 2026-11-16 11:38:10 +01:00
mesh-tools 97c8c81e82 Merge pull request 'Principle 10: the module is the unit, never the repository' (#623) from principles/module-is-the-unit into main 2026-11-16 10:35:37 +00:00
jschoubben b5e5a35460 Principle 10: a merge moves only the modules it changes 2026-11-16 11:35:33 +01:00
jschoubben 3c0389f74e Principle 10: the module is the unit, never the repository 2026-11-16 11:32:32 +01:00
mesh-tools 99e2966bb7 Merge pull request 'Design 47 G: the lead's answer A-447c (nothing is ever repository-based)' (#622) from design/47-G-lead-answers-40 into main 2026-11-16 10:31:45 +00:00
jschoubben 42054c599c Answer A-447c: nothing in building, checking or delivering works on a repository 2026-11-16 11:31:35 +01:00
mesh-tools f887e777b2 Merge pull request 'Design 47 G: the lead's answer A-447b (a build is always for one module)' (#621) from design/47-G-lead-answers-39 into main 2026-11-16 10:30:14 +00:00
jschoubben bc9ba62959 Answer A-447b: a build, a step and a check are always for one module; a fetch with no dir is refused 2026-11-16 11:30:04 +01:00
mesh-tools 4353e7c018 Merge pull request 'Design 47 G: the lead's answer A-447a (a module whose directory holds no go.mod; an empty choice fails)' (#620) from design/47-G-lead-answers-38 into main 2026-11-16 10:15:43 +00:00
jschoubben ca8cda69d9 Answer A-447a: a module whose directory holds no go.mod takes the go.mod files under it; an empty choice fails 2026-11-16 11:15:34 +01:00
mesh-tools a7b72d49d2 Merge pull request 'Design 47 G: the lead's answer A-447 (a Go fetch asked with a module's directory resolves that module alone)' (#619) from design/47-G-lead-answers-37 into main 2026-11-16 10:00:19 +00:00
jschoubben 2c399ac708 Answer A-447: a Go fetch asked with a module's directory resolves that module and its local replaces alone 2026-11-16 11:00:04 +01:00
mesh-tools 8c307e0e7a Merge pull request 'Design 47 G: A-446a, what goes with the runner's set-aside' (#618) from design/47-G-lead-answers-36 into main 2026-11-16 09:35:06 +00:00
jschoubben 2d6da3f34d Answer A-446a: a placed pipeline says pipeline-started; what only the set-aside used goes 2026-11-16 10:34:59 +01:00
mesh-tools 45b43ce81c Merge pull request 'Design 47 G: A-446, the pipeline runner holds no pipeline for its own new version' (#617) from design/47-G-lead-answers-35 into main 2026-11-16 09:25:12 +00:00
jschoubben 13ad18268f Answer A-446: the pipeline runner holds no pipeline for its own new version (the operator) 2026-11-16 10:25:02 +01:00
mesh-tools 5be43e0193 Merge pull request 'Principles: no part without a need (7); depend on the contract, at the level it is interchangeable (9)' (#616) from principles/contract-and-no-part-without-need into main 2026-11-16 09:06:33 +00:00
jschoubben 4e0f340916 Principles: 7 becomes 'no part without a need', 9 'depend on the contract, at the level it is interchangeable' (the operator, 2026-11-16) 2026-11-16 10:06:25 +01:00
mesh-tools 12c0f66fab Merge pull request 'After G: the module graph's edges are dependencies, never verbs' (#615) from design/47-G-after-delivery-2 into main 2026-11-16 08:57:25 +00:00
jschoubben 6ccd7c3edf After G: the module graph's edges are dependencies on a seat or a module, never verbs or bus subjects (the operator) 2026-11-16 09:57:18 +01:00
mesh-tools ea02e537a3 Merge pull request 'Design 47 G: the lead's answer A-445 (the bus keeper's start path)' (#614) from design/47-G-lead-answers-34 into main 2026-11-16 08:52:04 +00:00
jschoubben 60872e67e9 Answer A-445: the bus keeper's A-376 path and its four manifest lines go 2026-11-16 09:51:57 +01:00
mesh-tools 9cf4d63c9c Merge pull request 'Design 47 G: what is owed once G is delivered' (#613) from design/47-G-after-delivery into main 2026-11-16 08:46:19 +00:00
jschoubben 1a2d31b82f Keep what is owed once G is delivered in hq, not in a session's memory 2026-11-16 09:46:13 +01:00
mesh-tools 267f1ef0de Merge pull request 'Design 47 G: the lead's answer A-443a (a given-up publish step)' (#612) from design/47-G-lead-answers-33 into main 2026-11-16 08:42:19 +00:00
jschoubben 23cba849e3 Answer A-443a: a given-up publish step's call ends at the SDK's own bound and is said 2026-11-16 09:42:12 +01:00
mesh-tools 3723b2ea41 Merge pull request 'Design 47 G: the lead's answers A-394 and A-434 to A-444 (the start fixes' review)' (#611) from design/47-G-lead-answers-32 into main 2026-11-16 08:38:26 +00:00
jschoubben 62e3715af4 Answer A-394 and A-434 to A-444: the start fixes' review questions, each against a principle or the operator's word 2026-11-16 09:38:16 +01:00
mesh-tools 18ce04f6e8 Merge pull request 'Design 47 G: the lead's answers A-363 to A-372 (G starts)' (#609) from design/47-G-lead-answers-31 into main 2026-11-16 08:24:38 +00:00
jschoubben 91f60c7075 Use the glossary's words in lead answers 31, so the words test passes 2026-11-16 09:24:19 +01:00
jschoubben 4169ded5cd Answer A-433: before the catalog half is read, a module account's call runs unmarked, seat verbs too 2026-11-16 02:34:22 +01:00
jschoubben d9dd5dab67 Answer A-432: a runtime of its own runs the tool runner's program, which §524's check of the module's parts leaves out 2026-11-16 01:00:35 +01:00
jschoubben 2829a47ea0 Answer A-431: a take-in names the source revision it makes, one above the newest seen or said 2026-11-15 17:51:01 +01:00
jschoubben 09cf7d609d Answer A-430b: a branch build is named by the version the builder reserved; build.ended carries it and the registry tags it unreleased 2026-11-15 16:48:29 +01:00
jschoubben 547eec3373 Answer A-430a: every wanted build has a version; the node-engine resolves artifacts by version through the registry's versions verb 2026-11-15 16:44:17 +01:00
jschoubben d2094f657e Answer A-430: a want names the module version; the node-engine fetches the artifacts by that version (the operator) 2026-11-15 16:43:23 +01:00
jschoubben ee80b4d84b Answer A-426a: every rule waiting on a serving part counts any healthy running build that serves it 2026-11-15 16:17:01 +01:00
jschoubben 14c3e90332 Answer A-429: a released build is done for rule 1 2026-11-15 16:13:29 +01:00
jschoubben cac334e749 Narrow A-428: only the pipeline runner's own want skips the passed-pipeline gate while no runner runs 2026-11-15 16:04:05 +01:00
jschoubben aad8f7c921 Answer A-428: with no pipeline runner running, wants do not wait for a passed pipeline 2026-11-15 16:03:56 +01:00
jschoubben b20b382cd9 Answer A-426 (an acting wait ends on any running build that serves the edge) and A-427 (a put may hold the running build) 2026-11-15 15:56:51 +01:00
jschoubben 8abbb539fd Answer A-425a: which proposals are asked is the assigner's config key ask-operator-for 2026-11-15 15:55:34 +01:00
jschoubben f988c23d1e Answer A-425: a passing placement proposal is applied without the operator's ask (the operator, 2026-11-15) 2026-11-15 15:54:46 +01:00
jschoubben fd45f01929 Answer A-424: the operator channel keeps no limit of open asks per asker (the operator, 2026-11-15) 2026-11-15 15:49:03 +01:00
jschoubben f0b9c73a9f Answer A-423: one reader per durable through the hub; a consumed kept event with no route is said 2026-11-15 03:52:41 +01:00
jschoubben 17ab2d028e Answer A-422: the SDK's event hub binds the kept seat streams in its own start 2026-11-15 03:48:47 +01:00
jschoubben 9a60a312a0 Answer A-421: a trunk entry with no fp-src is matched to its release by the digests its build made 2026-11-15 03:17:45 +01:00
jschoubben 1de67e76ed Answer A-419 (release matched by the fingerprint's source) and A-420 (a message given up is said) 2026-11-15 03:14:06 +01:00
jschoubben 1088d86e5c Answer A-418: a bundle's or archive's name for reserved is <module>/<artifact> from its entry's artifact reference 2026-11-15 02:59:19 +01:00
jschoubben 2ce9268504 Answer A-417: a registry holder tags a version on the artifact it holds by digest, else takes it from the artifact store 2026-11-15 02:37:16 +01:00
jschoubben 39ab7efb6b Answer A-416a: the builder reads a registry seat's holder from the catalog's seat holders, not FollowSeat 2026-11-15 02:17:33 +01:00
jschoubben 6c5f8aabb0 Answer A-416: the builder knows a registry seat's holder through FollowSeat, not from running events since its start 2026-11-15 02:16:18 +01:00
jschoubben 3d56f25942 Answer A-415: a published heard before the build's digests is kept on its reservation and marks it when the digests land 2026-11-15 02:05:05 +01:00
jschoubben 9da181e12d Answer A-414: the grants part composes a module whose running build has no entry from its newest entry, as A-407 does for the node 2026-11-15 00:23:25 +01:00
jschoubben 8b9b450855 Answer A-413: named asks are unused only while no account the part composes may publish one 2026-11-15 00:03:35 +01:00
jschoubben 16f4fd8696 Answer A-412a: the entry's digest readings live in the mesh-catalog contract 2026-11-14 23:44:50 +01:00
jschoubben ee543b8601 Answer A-412: the builder takes the entry's manifest digest bare, and a look that writes nothing says why 2026-11-14 23:44:11 +01:00
jschoubben 7981d72d9b Answer A-411: the builder reads the last versions-filled from the seat event's kept stream, not EVENTS 2026-11-14 23:37:07 +01:00
jschoubben 842071859a Answer A-410: a running module the node cannot fingerprint is held, never counted as removed 2026-11-14 23:14:41 +01:00
jschoubben 11054da240 Answer A-409: a consumed seat event matches by name and filters on the seat's subject; the revision's id names its branch 2026-11-14 22:51:41 +01:00
jschoubben 46b6b74148 Answer A-408: the builder reads the kept versions-filled at its start 2026-11-14 22:47:32 +01:00
jschoubben 706e76b566 Answer A-407: a running build with no entry is described for resolving by the entry of the build its want says is being built 2026-11-14 22:46:19 +01:00
jschoubben 4b2eb0bcf4 Answer A-405 and A-406: a node composes around a running build with no entry; the pipeline's manifest passes the judge and is built once by hand 2026-11-14 22:42:00 +01:00
jschoubben 634b4ea49e Answer A-404: a source the gate started from the forge names no group in its source-moved 2026-11-14 22:16:04 +01:00
jschoubben b6380d0958 Answer A-403: the gate adds a registered repository a held source key lacks, and changes no head it holds 2026-11-14 22:08:40 +01:00
jschoubben dbe9d711ea Answer A-402: a head is read whole on any failure to read its merged message 2026-11-14 22:08:11 +01:00
jschoubben de9f6d3353 Answer A-401: a newly watched branch starts from the forge at its head, as a clone does 2026-11-14 22:05:17 +01:00
jschoubben d5e7d42d17 Answer A-400: seat-holders names the assigned claimant of a seat no assignments holds 2026-11-14 20:21:29 +01:00
jschoubben 232454cf76 A-397: the rule is the developer's judgement (share a durable only for shared work), not an operator ruling 2026-11-14 18:51:53 +01:00
jschoubben f55a6e7927 Answer A-399: a seat pulled by kind's holder says the seat's own events 2026-11-14 18:39:47 +01:00
jschoubben f6f82721d5 Answer A-398: a seat-form read resolves to the assigned claimant when no holder is on record 2026-11-14 18:17:34 +01:00
jschoubben 98e8bc395b A-397: the operator's decision time read from the clock 2026-11-14 18:03:19 +01:00
jschoubben ed95207361 Answer A-397: every process that hears events has a durable consumer of its own, named constantly; a shared one split the first registrations between the assigner's program and its tool runner 2026-11-14 18:02:11 +01:00
jschoubben 525a9590cd Answer A-396: every module is composed as invoking the tool runner's own reads
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 12:09:52 +01:00
jschoubben 6bc258db0d Answer A-395: every trunk part and the tool runner may read the catalog half
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 12:00:53 +01:00
jschoubben 5f04a5e3af Answer A-393: a node seat's verb asked with no machine is answered by any one holder
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 11:31:09 +01:00
jschoubben 087e17c799 Answer A-392: a seed asks a silent build seat once, and says so
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 11:24:49 +01:00
jschoubben 19f5ad86d2 Answer A-391: the indexer's start replay writes the view once, and says its length
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 04:03:00 +01:00
jschoubben 5d7a96fe66 Answer A-390: the verb revision's callers count an invokes of a single *, as grants reads it
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 03:19:13 +01:00
jschoubben 5aa4510f0f Answer A-389: before the catalog half is read, a module's call of another module's tool runs unmarked, logged
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 02:36:55 +01:00
jschoubben c1a647e9bb Answer A-388: the start path takes only an explicit read class; the broker's four reads are classed
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 02:22:16 +01:00
jschoubben 8dffde7f33 Answer A-387: before the catalog half is read, a read verb of a held definition runs; git is bootstrapped
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 02:20:39 +01:00
jschoubben 240f6401ce Answer A-386: a runner kept as it runs still gains the runner grants of what it carries
mesh/delivery ready: it is merged on the Approve
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 01:32:07 +01:00
jschoubben 77a6a9aa68 Answer A-385: a kept account is the union of the list in force and this pass's composition
mesh/delivery closed: a newer head of the same pull request
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 01:06:35 +01:00
jschoubben 9e619006c7 Answer A-384: a runner may make the ordered consumers its bundles hear new events on
mesh/delivery closed: a newer head of the same pull request
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 00:43:18 +01:00
jschoubben 8af69b2082 Answer A-383: a right every account carries reaches a kept account too
mesh/delivery closed: a newer head of the same pull request
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 00:22:48 +01:00
jschoubben 2c67b2e53b Answer A-382: every account may list stream names, which hearing an event starts with
mesh/delivery closed: a newer head of the same pull request
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-14 00:06:50 +01:00
jschoubben be264026b5 Answer A-381: grants' own account may list stream names, which its stream assertion starts with
mesh/delivery closed: a newer head of the same pull request
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-13 23:59:02 +01:00
jschoubben a95c0d6d98 Answer A-380: the runner's old-objects fallback uses the name grants creates
mesh/delivery closed: a newer head of the same pull request
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-13 23:18:26 +01:00
jschoubben c1065d9868 Answer A-379: the runner reaches a module's old objects by the name its membership gives, beside bindings
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-13 23:15:17 +01:00
jschoubben 0e9b6e5286 Answer A-377 and A-378: first-view tools pass the start path; the keeper may read the indexer's views
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-13 23:12:47 +01:00
jschoubben 64b9ffe8c1 Answer A-376: the keeper makes the catalog holder's own bus state before its catalog half is in
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-13 23:01:04 +01:00
jschoubben c5adb7fe48 Answer A-375: the gate's start path also lets the bindings' first view through
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-13 22:46:43 +01:00
jschoubben 61149124f8 Answer A-374: the seat verb withheld reads a small key beside the list
mesh/delivery closed: a newer head of the same pull request
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-13 22:10:26 +01:00
jschoubben 591297f02e Answer A-373: grants keeps its user list and pass in an object store, since they outgrow one bucket value
mesh/delivery closed: a newer head of the same pull request
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-13 22:00:46 +01:00
jschoubben e9a2aaf253 Answer A-371 and A-372: grants counts a seat stream it creates itself as empty, and the assigner asks Ready
mesh/delivery closed: a newer head of the same pull request
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-13 20:54:05 +01:00
jschoubben 904f5bfac5 Answer A-370: the indexer answers the catalog's first views at once, never behind a running take-in
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-13 20:53:47 +01:00
jschoubben da161cdbf1 Answer A-369: grants' catalog wait asks the client's Ready, as every part does
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-13 20:45:09 +01:00
jschoubben 916be3a63c Answer A-363 to A-368: G's first start composes rights from the trunk, builds unpublished packages at the source commit, and waits for nothing only G answers
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
2026-11-13 20:34:47 +01:00
5 changed files with 923 additions and 4 deletions
+4
View File
@@ -26,6 +26,10 @@ questions are.
- an interaction the design does not place that is not a mesh path: one module naming another, reading another's
internals or being special-cased in another part; a module's name in shared code; state only in memory; a poll or
a timer where an event exists;
- a value written into the code that can differ between machines, modules or over time (a port, a path, a bound, a
cap, a name, a time) and is not a setting with a default (principle 11); name the setting it should be, eagerly;
- a module that writes another module's file or state, or anything in another module's place, instead of giving it
through a provision or a contribution that module declares (principle 12);
- an effect or a reader of the build map's `What exists today` whose fate the diff does not carry out (run its
searches again on the reviewed tree), or one the search finds and the answer omits;
- a test that does not fail without its change (check it by taking the change out in a copy), an acceptance
+3
View File
@@ -25,6 +25,9 @@ write.
What you judge:
- Seams: where each piece of state lives on the bus (key-value bucket, stream, object store) and which module owns it; which seat, verb or event joins two parts; any link that is not NATS-native (a poll, a timer, HTTP between modules, a shared file, state only in memory); whether a form the manifest lacks is named with its cost (a mesh-sdk change rebuilds every module).
- Settings and ownership: every value that can differ between machines, modules or over time is a setting with a
default, never hard-coded (principle 11), and the design names it; no module writes another module's file or state:
what one module needs from another goes through a provision or a contribution that module declares (principle 12).
- Tools: for every new piece of state a tool that reads it, and for every new behaviour a tool that steers it (do by hand, ask again, drop, repair, re-check), as the instruction tools-are-how-a-module-is-spoken-to says.
Judge only whether the settled direction or design is complete and sound in your two architectural domains,
+12 -3
View File
@@ -1,6 +1,6 @@
---
status: canonical
updated: 2026-11-11
updated: 2026-11-16
---
# The principles
@@ -22,9 +22,18 @@ add, sharpen or drop a principle in a pull request to this file.
6. **Nothing is exposed publicly.** No data leaves the mesh unasked: no secret in a record, a log or a public
repository, no service reachable from outside unless it is meant to be. Inside the mesh, every agent and module
is trusted; who may do what is not a question to ask.
7. **The simplest thing that works.** Prefer what already exists. No new mechanism unless it is needed, and fewer
parts over more.
7. **No part without a need.** Every mechanism, field, seat and step exists because something needs it; prefer what
exists over a new one, and remove what no longer serves. Never a shortcut instead of the right shape.
8. **One word per thing**, from the [glossary](glossary.md).
9. **Depend on the contract, at the level it is interchangeable.** A module depends on a seat where the mesh means
more than one module to be able to fill the need, and on a module where only that module does; never on another
module's internals.
10. **The module is the unit.** Everything is done per module, never per repository. A repository is only where a
module's files are kept; a merge moves only the modules it changes.
11. **Configurable, never hard-coded.** A value that can differ between machines, modules or over time (a port, a
path, a bound, a cap, a name, a time) is a setting with a default, never written into the code.
12. **A module writes only its own.** A module writes its own files and state alone; what another module needs from it
goes through a provision or a contribution that module declares, never a file written in its place.
How it is checked: the one code review of a milestone checks the code against these principles and the design's
shape; an answer to a develop question names the principle that settles it.
@@ -3,7 +3,7 @@ layer: to-be
status: in-progress
design: 47
milestone: G
updated: 2026-11-13
updated: 2026-11-17
decisions:
- "ADR 0667"
---
@@ -746,3 +746,754 @@ the principle that settles it, cited as "design 47 G A-n".
on a cycle, so a call word's edge on a cycle is left out like an invoked verb's.
- **A-362** (completes A-353; principles 3 and 7) The seat list a runner starts with (mesh-sdk's bootstrapped list)
holds `mesh-delivery` beside `mesh-catalog`, so `mesh-delivery.source` is admitted before the seat list is read.
- **A-363** (cycle C7; principles 7 and 2) While its catalog half is unread, grants composes each account from the
manifest the indexer's `modules` first view names for its module (the trunk's), in union with what its want's entry
gives, never from the list in force alone; so the first G parts get their G subjects before any of them serves.
- **A-364** (completes A-363; principle 7) Grants' own account carries the right to create each kept event stream of
the seat list (`SEAT_*`, such as `SEAT_git_branch-moved`) while the catalog half is unread, composed the same way.
- **A-365** (cycle C4, a silence of ADR 0573 §2/§4, which speaks of the latest published version; principle 7) A mesh
Go package no published version holds is built, for its importer, at the commit the gate's `source` names for its
repository, as an import of the same repository is; once a version is published, the latest published applies. The
indexer asks such a build with that commit, not on the version line.
- **A-366** (W5; principles 7 and 3) The indexer takes its own build from `MESH_FINGERPRINT` (mesh-sdk
process.OwnBuild) first, and asks the node-engine's `composition` only when it is unset.
- **A-367** (W22; principle 3) A planner of the catalog half (the bus keeper) treats a `none-yet` trunk as not yet
viewed: it plans nothing, deletes no plan and says no module orphaned until the first revision with entries.
- **A-368** (C6; principle 7) When neither the registrations nor the gate's `source` name a branch (a fresh mesh),
the catalog half counts the trunk as `none-yet`, once in the catalog client's `Ready` for every part, so the approval
gate is ready and `indexer_seed_again` runs. In this release, since G's genesis needs it.
- **A-369** (completes A-368; principle 7) Grants' own wait for its catalog half (`CatalogOf.Waiting`) asks the catalog
client's `Ready`, so a mesh that names no branch counts as `none-yet` for grants as for every other part.
- **A-370** (cycle C8, found by dry run 2's baseline; principles 1 and 7) The indexer answers the first views of the seat
`mesh-catalog` (`registrations`, `revision`) at once from what its state holds, never held behind a running take-in;
a take-in that changes them says so on its events. So a runner's catalog half is read before the take-in's own calls
(such as the forge bridge's `bridge_merged_message`) reach it.
- **A-371** (review r1 finding 1; completes A-364; principles 7 and 3) A first view grants reads from a seat stream that
does not exist yet, and that grants itself creates (such as `SEAT_node-engine_running`), is empty, not a wait: grants
composes the pass, writes its own right to create the stream with it, then creates the stream and hears it from its
start; its health says the stream was created after the pass.
- **A-372** (completes A-368; principle 7) The assigner's wait for its catalog half (its `adopted` callback) asks the
catalog client's `Ready` as well, so a mesh that names no branch is `none-yet` for the assigner too.
- **A-373** (live start, novox 21:56: "the value for list in mesh-grants_user-list is 422584 bytes, over the 262144 bytes
the bucket takes"; principles 1, 3 and 7) Grants keeps its user list and its pass result, which outgrow one value of a
key-value bucket, as objects in an object store of its own state (declared in its manifest like any kept data), the key
naming the object; a write that cannot be made stays loud as today.
- **A-374** (completes A-373; principles 3 and 7) The seat verb `withheld`, served from grants' tools bundle, reads a
small key `withheld` that the process writes beside the list in its state `user-list` each pass it changes, so it
answers without the object the list became; before that key is first written it reads the list as before.
(Built in mesh-delivery d6d6bd18.)
- **A-375** (live start, novox 22:28-22:44: grants "the first view seat:mesh-broker.list of the bindings is not answered
(Not run: the marks of nats are not read yet ...)", and the indexer runtime "was bound no object-store called
catalog-revisions ... bound: none"; extends A-359; principle 7) The gate's start path also lets through the seat
`mesh-broker`'s verb `list` (the bindings' first view) while the catalog half is unread, since the catalog half's own
owner is bound its bus state through it.
- **A-376** (live start, novox 22:4x-23:0x: the indexer runtime "was bound no object-store called catalog-revisions";
the keeper plans nothing "before the catalog half's first views", which need that very store; cycle C12; completes
A-367 as A-363 completes grants; principles 7 and 3) Before its catalog half's first views are in, the bus keeper
makes the bus state of the holder of the seat `mesh-catalog` (the catalog half's own source), read from that module's
manifest on the trunk as the indexer's `manifests` first view gives it, and says each made entry as it does today, so
grants binds it; every other module waits for the catalog half as A-367 says.
- **A-377** (live start, novox 22:26: grants' first views of the indexer's `modules`, `manifests`, `verdicts` "Not run:
the marks of mesh-delivery-indexer are not read yet"; the keeper's A-376 reads the same views; widens A-359 back toward
A-353; principle 7) While the catalog half is unread, the gate's start path also lets through each module's own
first-view tools (the tools its first views name, which a caller list already limits), besides the seat verbs of A-359
and A-375.
- **A-378** (completes A-376; principle 7) The indexer's first views `modules` and `manifests` name the bus module (`nats`)
among their callers, so its keeper reads the catalog holder's manifest.
- **A-379** (replaces A-376 and A-378 for the start: the indexer's `catalog-revisions` is declared in its manifest's
old `objects` section, which grants creates (records.go objectsOf, assertions.go), while the tool runner reaches an
object store only by a new-form binding (`node-tools` internal/bus/bindings.go objectStore), so nothing binds it;
principle 7: what exists) The tool runner reaches a module's object store by the binding when one is bound, and
otherwise by the old `objects` entry its membership names, under the bus name grants created for it
(`OBJ_<module>_<name>`); a name neither names stays refused as today.
- **A-380** (corrects A-379's source of the name: grants issues no `objects` map in a membership; principle 7) With no
binding of that local name, the runner reaches the store under the name grants creates for a module's old `objects`
entry, `<module>_<name>` (stream `OBJ_<module>_<name>`); the bus's grants to the module's account decide whether it
may, and a refusal is said with the store's name.
- **A-381** (live start, novox 23:42-23:58: grants is "Permissions Violation for Publish to $JS.API.STREAM.NAMES" while the
loaded list grants it STREAM.CREATE/INFO of each kept seat stream, so its assertion of the seats' kept events never
reaches the create and SEAT_mesh-conditions_findings-asked stays missing; every bundle with a finding writer exits;
principle 7) Grants' own account carries `$JS.API.STREAM.NAMES` (the read its assertion of kept streams starts with),
composed with the create rights of A-364.
- **A-382** (widens A-381 to its kind, met a second time: live 00:02, every G bundle with a finding writer exits "no stream
on the bus carries mesh-conditions.findings-asked", because the tool runner's SubscribeNew finds a stream by subject
through `$JS.API.STREAM.NAMES`, which `novox.node-tools` may not publish, while EVENTS carries `mesh.seat.*.event.>`;
principles 7 and 6) Every account grants composes carries `$JS.API.STREAM.NAMES`, the read of stream names by subject
that hearing an event starts with; it lists names only.
- **A-383** (completes A-382; live 00:21: list revision 274 loaded, yet `novox.node-tools` still lacks
`$JS.API.STREAM.NAMES` (1805 publish rights, unchanged), since grants keeps an account it holds "beyond this pass" as
the list in force while the catalog half is unread; principles 7 and 3) A right every account carries (A-382) is added
to every account grants writes, a kept one included: keeping an account in force never withholds what every account
is given. A pass that only adds still never takes from a loaded user.
- **A-384** (completes A-382/A-383 for the whole hearing path, met a third time: live 00:35 "an ordered consumer of
mesh-conditions.findings-asked on EVENTS: context deadline exceeded", the runner account holds only INFO of named durables
on EVENTS; principles 6 and 7) Every tool runner account carries what the runner's SubscribeNew (an ordered consumer from
the next message, mesh-tools `node-tools` internal/bus/subscribenew.go) asks of the stream that carries the events its
bundles hear: the create of an unnamed consumer filtered to those events' subjects, its next-message fetch, its info and
its delete, on EVENTS and on every kept seat stream; the same form for every runner, composed in one place.
- **A-385** (generalizes A-383, met a fourth time: live 01:00 "mesh-delivery-gate's bundle asked mesh/ledger.get and it
failed after 10s", the gate's ledger `records` exists and is bound, and `novox.node-tools` holds no right on
KV_mesh-delivery-gate_records_state, since a held runner account is copied whole from the list in force; principles 7
and 3; A-363's union) An account grants keeps beyond a pass, a runner's included, is written as the union of the list
in force and what this pass composes for it (its carried modules' grants from the trunk's manifests and their
bindings), never the list in force alone; a pass that only adds still never takes from a loaded user.
- **A-386** (completes A-385 for runners kept as they run; live 01:31: grants 3c582b1b in, `novox.node-tools` unchanged
(2004 publish rights, none on KV_mesh-delivery-gate_records_state), since `runnerUsersKept` keeps a runner whose running
manifest does not say it is the tool runner whole (ADR 0601 §4); principles 7 and 3) A runner kept as it runs is written
as the union of the list in force and the runner grants its carried modules' bindings and events give (the ledger
readers, the new-event rights of A-384, the streams' names of A-382): what a carrying runner is given, never what a
plain module is given, so ADR 0601 §4 stands.
- **A-387** (generalizes A-353, A-359 and A-375, met a fourth time: live 02:16 the first take-in took in none of 18
repositories, "the seat git did not answer pull: Not run: the definition of the seat git is not yet read by this tool
runner's seat list", and the runner's own `mesh-catalog.seat-definitions` is refused "the marks of
mesh-delivery-indexer are not read yet"; principles 7 and 3) Before the catalog half is read, the gate lets through
every verb that the seat definition the runner holds classes `read` and does not mark needs-approval, and judges every
other call as before; the hand-kept list `startReads` goes. The bootstrapped definitions carry every seat the first
take-in reaches: `git` joins them, a byte-for-byte copy of its contracts file with its row in the bootstrap test. The
catalog seat's contract classes `modules`, `module`, `provides`, `dependents` and `seat-definitions` as `read`, and
its bootstrapped copy follows.
- **A-388** (completes A-387; the developer's question: seatdef callers.go reads a verb with no class as a read, which
would let unclassed acts such as `mesh-delivery-checker.try` and `operator-channel.tell` through before the catalog
half is read; principles 3 and 7) A-387's start path takes a verb whose definition names the class `read` explicitly,
never one that names none. The broker seat's contract classes `list`, `show`, `check` and `orphans` as `read`, and its
bootstrapped copy follows, so A-375's `mesh-broker.list` still passes.
- **A-389** (completes A-387 for module tools, met a fifth time: live 02:32 the take-in's
`mesh-delivery-forge-bridge.bridge_merged_message` is refused "the marks of mesh-delivery-forge-bridge are not read
yet"; a module tool names no class, and its marks live only in the catalog half's entry the take-in has not made yet;
principles 7 and 3) Before the catalog half is read, a module tool whose proven caller is another module's account (never
a session's or the mesh MCP server's) runs unmarked, and the runner logs each such call once per tool with its caller ("run
unmarked before the catalog half is read"); a module tool called otherwise waits and is judged as before. A module
tool's mark is read again, and holds, from the call after the half is read. Today the one marked module tool,
`gitea_branch_protection_loosen`, has no module among its callers.
- **A-390** (live 03:17 on novox: the tool runner's first view of the catalog half is refused "mesh-catalog.revision is
refused to `novox.node-tools`: the tool answers only mesh-delivery-builder"; `node-tools` declares `invokes: ["*"]`, and
the verb revision's callers rule (ADR 0663 §2) matches `seat:mesh-catalog.revision` by its exact text alone, while the
grants part reads a single `*` as every seat verb; principles 7 and 3) The verb revision's callers are the parts whose
entry in the newest trunk revision declares `seat:mesh-catalog.revision` or a single `*` in invokes, as the grants part
reads invokes; the holder stays off the list.
- **A-391** (live on novox: the indexer's start ran 50 minutes of CPU in its first hour without reaching its take-ins
watch, so the seed was never asked; it wrote the whole catalog view again after each finished build of the build log,
though none changed a module; principles 7 and 3) A finished build writes the view again only when it changed a module;
at the start the build log's finished builds are applied first and the view is written once, the replay said in one
line at its start and one at its end with its count and duration.
- **A-392** (live on novox: the seed asked every artifact's build seat for its steps one by one, each call to a seat with
no holder retried for 30 seconds, so a seed of 178 entries ran for hours before refusing; principles 3 and 7) Within one
seed, a build seat that did not answer is not asked again for its later artifacts; they stay unanswered, the seed refuses
with the same words, and the seat's first silence is said in one line.
- **A-393** (live: the build runtimes serve `steps` only on `mesh.seat.node-build-go.tool.steps.<node>`, and the seed asks
`seat:node-build-go.steps` naming no machine, since it cannot know the holders before the assigner's view; principles 7 and 1)
A verb of a node-scoped seat asked with no machine is answered by any one holder: each holder's runtime also serves the verb
on the subject without the machine (tool and call forms), in one queue group per seat and verb, and the grants part grants
each holder those subjects.
- **A-395** (live: novox's tool runner was refused `mesh-catalog.revision` 54 times, the verb answering only parts that
declare it in invokes, while the runner reads the catalog half for every module it serves; principles 7 and 4) Every part
the newest trunk revision holds may read the verb revision, and so may the tool runner; the holder stays refused. ADR 0663
§2's rule that a part reads it only by declaring it is proposed for its end.
- **A-396** (live: the shanks build runtimes' seat list and catalog-half reads were refused on their caller-named subjects,
since no build module's manifest invokes them; principles 7 and 4) Every module is composed as invoking the tool runner's
own reads (the seat list's verb, the catalog half's revision and registrations, the delivery seat's source), so each
account the runner serves a module on may ask them.
- **A-397** (live, ace 2026-11-14: the assigner's program and its tool runner both bound the durable consumer
`ace_mesh-delivery-assigner` on `SEAT_mesh-catalog_registration-changed`; the runner pulled all 178 events of the first
registrations and the program, which acts on them, heard none and said nothing; principles 1 and 7) A durable name is
shared only where the work is meant to be shared: a work queue, each item to one taker, as NATS shares one durable among
its pullers. Every other reader needs every message, so each process of a module that hears events has a durable consumer
of its own, named constantly across restarts: the module's program `<node>_<module>`, its tool runner
`<node>_<module>_tools`, any other process its manifest names `<node>_<module>_<process>` (A-340). A durable name holds no
dot (design 19 *The consumer*). The grants part composes and makes each, with the hear of exactly that process's events.
This is a builder's judgement no design line overrode (the operator, 2026-11-14: "the developer should use his brains");
design 19's *one durable consumer per module* was written when a module was one process.
- **A-398** (live, shanks 2026-11-14 18:13: the G builder was refused `KV_mesh-build-go_steps` and the image and node steps,
since its reads `seat:node-build-{go,image,node}.steps` resolved to no holder: no assignments holds the build seats; the
operator decided, 2026-11-14 about 18:16: "yes, do it", after their word of 16:45 that who holds a seat follows from the
module graph and the assignments; principles 1 and 7) A seat-form read, view or view tool resolves to the holder on
record, else to the first module assigned to a node (nodes and modules in name order) whose manifest claims the seat.
Holds keep heldWherePlaced's rule. The whole question of holders stays with the post-G remarks.
- **A-399** (live, shanks 2026-11-14 18:32: the Go build seat was refused `toolchain-reported.shanks`, and `step-ended` too: a
seat pulled by kind is composed by the seat-traffic rules, which grant an event only under a kind or a caller's name, and
the held-seat rule skips such a seat, so its own events were granted to nobody and the builder heard no toolchain;
principle 7, and who-may-is-not-the-operators-question) The holder of a seat pulled by kind (not kinded) says each of the
seat's events no caller names: on its own machine in its account where the definition's subject names `<node>`
(machineEvent now reads `<node>` from any node seat's event subject, not the node-engine's alone), and with any machine in
its membership, since the runtime carrying it checks the membership; the account is therefore granted any machine's too.
- **A-400** (live, the control-node 2026-11-14 19:12: the indexer waited for ever on the view `seat:mesh-module-assigner.assignments`
for a holder of the seat, since the assigner's state seat-holders is written from the assignments' holds alone and no
assignments holds mesh-module-assigner; its versions fill stayed undone and no build began; the operator, 2026-11-14 about
16:45 and 18:16, as for A-398; principles 1 and 7) The assigner writes seat-holders from the assignments' holds and, for
each seat no holding names (a seat held once: on no node; a seat of every node: on that node), the first module assigned
to a node (nodes and modules in name order) whose manifest of the trunk claims it: the same rule as A-398, in the
writer every reader of the holders reads. It says each key so derived once, and when it stops being derived. The
assignment graph's requirement resolution and the order waits keep reading holders from the holds alone; the post-G
remarks take that up with the rest of the holders question.
- **A-401** (live, the control-node 2026-11-14 21:5x: the assigner ended every proposal, since f1-circuits (watched on
`master`), link2pay and photos (watched on nox-mesh) are held in no catalog graph of their branch; the gate writes
`source.<branch>` only on a take-in of a merge on that branch (A-34), so a newly watched branch is never started; the
operator, 2026-11-14 about 22:00: "by doing a git clone"; principles 2, 4 and 7, §15, §430, A-85, A-93) A watched branch
starts from the forge, as a clone does. The gate, at its start and on each `registration-changed` of the seat
`mesh-catalog`, writes `source.<branch>` for every watched branch the registrations name that no source key holds yet
(§430, which wrote the default branch alone): each repository registered on it at the branch's head, asked of the seat
`git`'s verb `branch`, the group left empty, and says `source-moved`; a repository whose head cannot be read is left out
and said, and is added at the next registration change or by hand. Its tool `gate_source_start` (branch, why) does the
same by hand through the hand-act log. A key it holds is never rewritten this way: from then on only a take-in moves it
(A-34). The indexer takes a head the source names in from that head's merged message as before; a head with no merged
message is read whole at that commit through the seat `git` (its tree and files), taken in as a merge whose changed
files are every file of that tree, and computed as any other (A-93). So the branch keeps its first revision
`branch.<branch>.1`, and its later merges follow the designed path.
- **A-402** (A-401; principles 7 and 3) The indexer reads a head whole through the seat `git` on any failure to read its
merged message, not only on the forge bridge's "no merged pull request": the read at the same commit gives the same
tree, so no typed error is added to the bridge; each such read is logged with its cause, and a head neither way reads
stays unread, as before.
- **A-403** (A-401; principles 2 and 7) "A key it holds is never rewritten" means no head it holds is changed: at its start,
on each registration change and by its tool, the gate adds to a held `source.<branch>` every registered repository the
key does not name yet, at the branch's head, and says `source-moved` only when it added one; the heads it holds move by
a take-in alone (A-34). So a repository left out once, or registered later on a branch already watched, comes in the
same way.
- **A-404** (A-401, A-403; principles 7 and 3) A `source-moved` the gate says for heads it started from the forge, on a new
key or added to a held one, names no group, as A-401's new key does: no group moved it, so nothing times it as a
group's; the key keeps the group a take-in last wrote.
- **A-405** (live, the control-node 2026-11-14 22:34: its node-engine refused every candidate, "nothing provides" postgres-database,
secret, route, s3-bucket, document-server and mcp-endpoint, since its catalog holds only modules whose wanted build has a
catalog entry, and the providers' wants name the builds running from before G, which have none, their new builds waiting
for a pipeline that cannot be placed while the node refuses; the operator, 2026-11-14 about 22:40: "yes, go ahead";
principles 2 and 7, §421) A module the node runs whose wanted build has no catalog entry stays in the node's catalog as
the build it runs, from the manifest of that build the node keeps: it provides, claims and requires what that manifest
says, so the node resolves around it. Its want stays held with `entry-missing` (§421) and nothing new of it is composed;
when its new build's entry comes, it is composed as any other.
- **A-406** (live, the control-node 2026-11-14 22:34: the pipeline's manifest at its build was refused by the manifest judge on
2026-11-13, and its candidate carried no parts; the operator, as for A-405; principles 5 and 7) The pipeline's manifest is
made to pass the validator, as any module's; its first build is asked of the builder once by hand, the way a run on request
asks it, since only a pipeline asks builds and none runs yet.
- **A-407** (A-405; principles 7 and 2: the node keeps no manifest of a build that has no entry, and the running event
carries only a digest) A running module whose wanted build has no entry is described from the catalog entry of the build
its want says is being built (the newest of its first watched branch), read through the seat `mesh-catalog`'s `entry` as
any entry: what it provides, claims and requires there stands for the running build, for resolving the node alone. A
want that names no newer build leaves the module as today (a name alone), said once.
- **A-408** (live, the build node 2026-11-14 22:45: the builder held every build "wait-held, the catalog has not said
versions-filled", while the indexer said it once at 01:33 UTC and keeps it as said; the builder keeps the mark in memory
alone and its durable consumer took the event once, so a restart lost it; principles 1 and 3, §369, §372) The builder
reads at its start the last `versions-filled` the seat's stream keeps (it is kept with no age limit, §372) by a direct
read of the last message on its subject, and takes it as heard; the event heard later sets it as before.
- **A-409** (live 2026-11-14: the assigner's and the builder's durable consumers filter only
`mesh.mod.mesh-catalog.event.catalog-revision`, so neither hears a catalog revision after its first view; the seat
definition says the event `catalog-revision` with the subject `catalog-revision.<branch>`, and the grants part, finding
no match for the consumed `mesh-catalog.catalog-revision`, fell back to a module's event without a word; principles 1
and 3) A consumed seat event matches the seat's event of that name whatever its subject's further tokens, and its filter
is the seat event's subject with each `<...>` token as `*` (a trailing one also as the rest); a consume naming a seat
that says no such event is refused aloud by the grants part, never turned into a module's event. The indexer's
`catalog-revision` duplicate id names the branch, and it is published with the bus's acknowledgement.
- **A-410** (live, the control-node 2026-11-14 about 22:55: proposals placing the pipeline were answered "differs", with
fail2ban, pacman and mesh-mail-reader counted as removed, while all three run and their assignments hold them; a module
started before the node-engine wrote fingerprints has none in what it runs, and none can be made for it, so its
comparison is skipped, and a skipped comparison was counted as a removal; principles 2 and 3) A module the node runs
whose fingerprint it cannot read or make is held: it is kept as it runs and said as held, with why, in the node's
answer, never counted as removed and never stopped for it. Only a module the node runs that no assignment names is
removed.
- **A-411** (A-408; live, the build node 2026-11-14 23:27-23:34: the builder's read of the stream `EVENTS` is refused, since
no module is granted that stream's details, by design; `versions-filled` is a kept seat event on its own stream, which the
builder consumes and is granted; principles 1 and 7) The builder reads the last `versions-filled` at its start from the
seat event's own kept stream, through the SDK's hearing of a kept event read from its first message, as any kept event
is heard; it reads nothing of `EVENTS` and needs no new grant.
- **A-412** (live, the build node 2026-11-14 23:40-23:41: the builder's claim of every build is refused, "is not a manifest
digest, 64 lowercase hex digits", since the indexer's catalog entry now says its manifest digest as `sha256:` and hex,
and the builder passes it on as it reads it, while its builds and its manifests are keyed by bare hex; a build asked
again then answers "wrote nothing … its log says why" with no why; principles 3, 7 and 8) The builder takes a catalog
entry's manifest digest in its bare form, through the contract's own reading of it, wherever it plans, claims or keys
a build; the form the entry says is the entry's. A look that writes nothing says why in the builder's log, for each
of its ends, and the answer to a build asked again carries that why.
- **A-412a** (A-412; principles 1 and 7: the form of a seat's entry is its contract's) Both readings of an entry's
manifest digest (the entry's form and the bare one) move into the seat `mesh-catalog`'s contract package; the indexer
and the builder use them from there, and neither keeps its own.
- **A-413** (§544 as G-12 answered it, A-170, A-231; live, the bus 2026-11-14 23:49: the builder of G asked 113 builds on
`accept.delivery.<asker>`, and no named worker exists to take them, since the grants part counts named asks unused while
no entry of the revision invokes a build seat's queue, and in the start no entry of G is in the revision yet; principles
3 and 2) Named asks are unused only when no entry of the revision invokes a build seat's queue and no account the part
composes is granted to publish a named ask; while either holds, each build seat's queue keeps its named worker, made at
the part's next pass. The purge at a revert stays as §544 says, for a revision whose accounts ask no named ask.
- **A-414** (A-407, A-413; live, the bus 2026-11-15 00:17: after A-413 the grants part still made no named worker, since the
holders of the build seats, and the builder, run builds from before G with no catalog entry, so the part composes none of
them and carries their last memberships unchanged; the operator, 2026-11-15 about 00:20: "yes"; principles 2 and 7) The
grants part composes a running module whose wanted build has no catalog entry as A-407 describes it for the node: from
the catalog entry of the build its want says is being built (the newest of its first watched branch), read through the
seat `mesh-catalog`'s `entry` as any entry, for its account's grants, membership and seat traffic alone. A want that
names no newer build leaves its membership carried as today, said once.
- **A-415** (§588, A-73; read from the code 2026-11-15 by the developer, the live proof waits on its new log lines: the
builder records a trunk build's digests on its reservation only after the build's claim is finished, while the registry
publishes during the build, so a registry seat's `published` heard before then marks nothing, is acknowledged and is
lost, and package-published is never said, as for 0.1.6; principles 1, 3 and 7) A `published` that marks no artifact
of its reservation yet is kept on that reservation, in the builder's own kept state, as heard (name, version, digest);
when the build records its digests, each kept `published` whose digest matches marks its artifact and package-published
is said then, in the same way as for one heard after. Nothing asks the registry; a kept `published` whose digest never
matches is said in the builder's log when the build records its digests.
- **A-416** (§588, A-302, A-415; live, the build node 2026-11-15 02:02 and 02:16: `builder_publish_again` for
mesh-delivery-pipeline 0.1.6 queued nothing, "no module running that claims the registry seat oci-artifact-registry
is known to the builder since its start", since the builder learns a registry seat's holder only from node-engine
`running` events heard since it started, and that knowledge dies with every restart; principles 1, 4 and 7) The
builder knows each registry seat's holder through the SDK's FollowSeat, as A-302 has the indexer follow its seats,
from its start; a `running` heard still queues the versions waiting on that holder, as §588 says, but is no longer
how the holder is known. A publish again whose seat has no holder said by FollowSeat says so, naming the seat.
- **A-416a** (A-416, A-298; principles 1 and 7: one source of who holds a seat, and nothing followed that nobody reads)
FollowSeat follows a state of a holder, and the registry seats keep none the builder reads; so the builder knows a
registry seat's holder from the seat-holders reading A-298 names (`SeatHolders` of the catalog client), the one its
FollowSeat of the assignments already takes, read from its start. Its manifest and grants do not change. A-416 is
read with "the SDK's FollowSeat" as this reading.
- **A-417** (A-73, A-50, §588; live, the control-node 2026-11-15 01:57 and 02:28: the build node's publish of
mesh-delivery-pipeline 0.1.6 stores each artifact untagged, as A-73 has it, and the registry holds it, under the
artifact digest the builder reserved; the builder's publish of the version with that digest and no parts is then
refused, since the holder of `oci-artifact-registry` says published only for a tag it already holds and otherwise
opens a transfer with no parts; principles 3 and 7) A registry holder given a publish with a version, a digest and no
parts tags that version on the manifest it holds of that artifact digest under the name, and says published; when it
holds none, it takes the bytes from the artifact store by that digest, as A-73 and A-50 say, and stores and tags them;
when neither has it, it refuses as not stored, naming the digest. The builder's publish again names its transfer in
the transfer form the build's own step uses.
- **A-418** (§478, §676, §694, A-417; live, the control-node 2026-11-15 02:57: the tag of mesh-delivery-pipeline 0.1.6 is
refused since the builder's `reserved` answers "no entry carries the name mesh-delivery-pipeline/program": it matches a
name only against an artifact's `package`, which a bundle or an archive has none of, so no bundle is ever found;
principles 7 and 8) A bundle's or an archive's name for `reserved` is `<module>/<artifact>`, read from its catalog
entry's artifact reference (`<module>:<artifact>`), the name the build step and the builder publish it under; the
entry stays the one source of the name, and no `package` is added to it. A name that matches no entry is said in the
builder's log with the entries looked at.
- **A-419** (A-96, A-67; read from the code 2026-11-15: the assigner finds a trunk build's released version and parts by
its module and the source revision of the kept assignment graph, so after any merge that does not rebuild a module,
its wants and candidates carry no version and no parts; principles 2 and 7) A trunk build's release is matched by its
module and the source revision its fingerprint was built at, read from its catalog entry's `fp-src`, which is
`version-released.source` for that build; A-96's "matched to `version-released.source`" is read so. A want or a
candidate whose entry's release is not heard says so, naming the module, the fingerprint and that source.
- **A-420** (principle 3; the builder and the indexer hear their events with five deliveries, and the bus drops a message
after the fifth without either saying so) On the last delivery its consumer allows, the builder and the indexer each
say in their log that they give the message up, naming its subject, the artifact and version it carries, and the
last refusal. (Quick fix, said: the same silence holds for every part that hears an event; the thorough fix is the
SDK's hearing saying it once for every part, noted on the issue of the hand start.)
- **A-421** (A-419, ADR 0601 §1; read live 2026-11-15: six entries of trunk revision 225 carry no `fp-src`, since a
fingerprint that moves without a merge clears it, so A-419 finds no release for them; principles 2, 3 and 7) A trunk
entry with no `fp-src` is matched to the release whose parts' digests are the digests its fingerprint's `build.ended`
made, as the assigner already hears it; an entry with `fp-src` is matched as A-419 says. An entry neither finds is
said as A-419 says, with which of the two was tried.
- **A-422** (A-397; live, the bus 2026-11-15 03:45: the indexer, the assigner, the checker, the teller, the forge bridge,
the view, distribution and nats each have a durable on a kept seat stream their manifest consumes that was never
delivered, since only the builder binds its kept streams; the indexer never heard a package-published; principles 1,
3 and 7) The SDK's event hub binds a part's kept seat streams in its own start, after the stream `EVENTS`, so every
part hears each kept event its manifest consumes on its own durable there, as A-397 says, with no call of its own;
a stream it cannot bind is said in its log and its start goes on. The replay of what waited is heard as any event;
a part whose handler cannot take its backlog is a finding of the change, read before its delivery. (Thorough fix,
said: the indexer's own call, b0242f68, is the quick fix placed meanwhile.)
- **A-423** (A-397, A-422; read from the code 2026-11-15: the assigner reads machine-away and machine-back, and the forge
bridge branch-moved, with readers of their own on the durable the hub now binds too, so each message would reach one
of the two; and the checker, the forge bridge, the view and distribution consume registration-changed in their
manifests with no route, so its durable fills unread; principles 1, 3 and 7) One reader per durable: the hub is the
only reader of a part's kept streams, and a part that read one itself hands its handler to the hub, which gives each
event to every handler it has for it. A kept event a part's manifest consumes and no route takes is said at its start,
naming the event, as a fault of the part to be fixed in its manifest or its code.
- **A-424** (§394, §450, A-423; the operator, 2026-11-15 about 15:50 CET: "there is no limit for the amount of asks in my
telegram", said many times before; live: the teller held the pipeline's proposal ask behind two week-old asks, since
an asker may have at most three asks open; principles 3 and 7) The operator channel keeps no count of open asks per
asker and holds none back for room: every ask is offered when it is raised. The router's limit, the teller's own and
the ask-again for a held-back ask (§394, §450) go; an ask still lapses at its expiry or ends at its answer.
- **A-425** (the operator, 2026-11-15 about 15:57 CET: "yes, remove the per-placement approve", after the lead proposed
it under propose-removing-a-wrong-rule: a person's yes for every placement holds work on a form, against "merges and
deliveries need no approval"; principles 3 and 7) A proposal whose verdict passes is applied by the assigner at once,
with no ask of the operator, when it only places, pins, unpins, holds, unholds, decides or configures. A proposal
that takes a module off a node, removes a node, or whose verdict says a node would stop or remove a module or its data,
is still asked of the operator, as are the asks of a module whose moves a person takes. Each applied proposal is said
as an event and in the assigner's log, naming who proposed it and why. The decision record that made every proposal
wait on the operator's yes gets a new record saying so once the knowledge base admits callers again.
- **A-425a** (A-425; the operator, 2026-11-15 about 16:00 CET: "this is all configuration, a simple, single mesh mcp
call"; read live: the assigner declares only `outside-settle-seconds` and `outside-wait-minutes`, and asks every
passing proposal in its code; principles 4 and 7) Which proposals the assigner asks of the operator is the assigner's
declared config key `ask-operator-for`, a list of change kinds (`place`, `take-off`, `pin`, `unpin`, `hold`,
`unhold`, `decide`, `config`, `add-node`, `remove-node`, and `stops` for a verdict that stops or removes a module or
its data), set at mesh level with `mesh-config.config`; its default is A-425's: `take-off`, `remove-node`, `stops`.
A proposal none of whose changes is listed is applied at once. The asks of a module whose moves a person takes stay
as they are. A change of the key is said in the assigner's log at its next pass.
- **A-426** (§268, A-323, A-330; live, the build node and ace 2026-11-15 15:52-15:56: the pipeline's want on novox
waited for an assigner build not yet built, while a working assigner ran, and only a hand release moved it; the
operator: "then our design is stupid"; principles 2, 3 and 7) An acting wait waits only while its acting part runs
nothing that serves the edge: an acting part that runs a build serving the edge's seat or verb ends the wait at
once, whatever version its newest is, and the wait says which build it took. A wait on a version not yet built is
never held for that alone. (This replaces the node-engine exemption of A-323 and A-330 with a rule for every part.)
- **A-427** (A-426; live: the builder refuses a put of a version it did not build, so a part started by hand cannot be
held, and the trunk's newest build of it would replace it; principles 2 and 4) A put may name the build a node
runs now, read from the node's running statement, and holds the want there; the builder's store is not asked for it.
- **A-428** (§29, A-328; read from the code 2026-11-15 16:05: a want follows a passed pipeline, and no pipeline can
pass while no holder of the seat `mesh-pipeline-runner` runs, so the runner itself never gets a want; the gate
returned before the want rules without a word; principles 3 and 7) While no module holding `mesh-pipeline-runner`
runs on any node, a want of a module whose manifest claims `mesh-pipeline-runner` does not wait for a passed
pipeline: the assigner writes it from the trunk's entry by its other rules, and says once that it does and why.
Every other module still waits for a passed pipeline, so nothing else moves before a runner runs; once one runs,
A-328 has the runner's own changes pass on it. (Whether a want waits for a passed pipeline at all is a gate for the
generic gating research to make configured.)
- **A-429** (§379, A-421; live 2026-11-15 16:07: the assigner's rule 1 said a released build "is being built", since
it knew build ends only from a stream it read from its start on, kept by age, not one per subject as the builder's
manifest declares; principles 2 and 7) A build whose version is released (`version-released` heard for its
fingerprint's parts) is done for rule 1, as §379 takes its parts from that event; `build.ended` counts for a build
not released (another watched branch, a kept build). The assigner reads the builder's build ends whole at its start.
That the bus keeps that stream by age, not one per subject as declared, is a fault of the grants part, noted.
- **A-426a** (A-426, ADR 0614 §2; live 2026-11-15 16:15: rule 8 holds the pipeline runner's want, "the verb
seat:mesh-hand-acts.hand-act mesh-delivery-pipeline invokes is served by no running build", while mesh-hand-acts runs
on the control-node and serves that verb with a build from before G; principles 2, 3 and 7) A-426 holds for every
rule that waits on another part, the acting waits and rule 8 (an invoked verb served) alike: a part counts as
serving when a node runs a build of it that serves the seat or verb and is healthy, from its running statement and
the tool runner's announcement, whatever its version and whether or not the catalog lists that build. The rule says
which build it counted.
- **A-430** (§379, §525, §526; the operator, 2026-11-15 about 16:45 CET: "the artifacts simply carry the module
version of course, why would a module have a version otherwise"; principles 6 and 7) A want names the module and its
released version (and the kept-on node for a kept build), never its artifacts' digests: the node-engine fetches each
artifact the module's catalog entry names under that version from the registry seat that holds it, as the registry
tags it, and the registry's answer is what it applies. §379's digests in the want, and the assigner's check whether
a node's node-engine reads them (`want-parts`), go. A build of another watched branch, which has no released
version, is named by the version the builder numbered it under.
- **A-430a** (A-430; principles 7 and 8) Every build a want names has a version: a released trunk build its released
version, a build of another watched branch the version the builder numbered it under (A-430), and a kept build its
numbered version with `kept-on` naming the node that keeps it (§526); a want never names a fingerprint's artifacts
or digests. The node-engine resolves each artifact the catalog entry names at that version through the seat that
holds its kind (`oci-artifact-registry` for a bundle or archive, `container-image-registry` for an image), by its
verb `versions` with the artifact's registry name `<module>/<artifact>`, and fetches the bytes as today over the
artifact store by the digest that answer gives; an image is pulled by name and version. The node-engine's manifest
invokes those `versions` verbs; no registry verb is added.
- **A-430b** (A-430a, §52, §53, §371; principles 7 and 8) The version a build of another watched branch is numbered
under is the one the builder reserved when its build started (§52), never released: `build.ended` names it as
`version` for every build that reserved one, and the registry seat that holds each artifact tags a branch build's
artifact at that reserved version without releasing it, so the node-engine resolves it through `versions` like a
released one. §371's and §53's "stored by digest with no version" becomes "tagged at its reserved version,
unreleased"; the `package-published` comparison by digest stays. Order within G: the node-engine's resolution by
version first (it unblocks every released build), then the builder and registry tagging; until those land, a want
for a branch build keeps its digests for a node whose node-engine reads them, and that form goes with them.
- **A-431** (§476, §493, §648, A-34; live 2026-11-15 17:44: the pipeline runner heard eight `take-in.taken` with no
branch and no source revision, started no pipeline and asked each again for ever; principle 7) The source revision a
take-in names is one above the newest the indexer has seen of that branch (the gate's `source-moved`) or named in an
earlier `take-in.taken`; the gate's `source.<branch>` then moves to it (A-34). Every taken event of one take-in carries
the same `branch`, `source` and `entries`, so a reader keyed by branch and source revision (the runner, the gate)
acts once; a taken event that took nothing in (a group's end, one said before G) carries none, and the runner writes
it in its journal and lets it go (§493).
- **A-432** (§524, ADR 0259 §8; principle 7) A module of its own account is run by a runtime of its own, a process
whose program is the tool runner's build, not the module's: the check of §524 leaves that process out, and the tool
runner's own want checks its program. Every other archive, process and container of the module's composition is
checked against its want's `parts` as §524 says. (novox, 2026-11-16 01:00 CET: the new indexer was held
"parts-differ: fetched and not in the want's parts" for the tool runner's program.)
- **A-433** (A-389, §19; principle 6) While its catalog half is unread, a tool runner runs unmarked every call whose proven
caller is a module's account, a seat verb as a module tool, and a module's own call in a runtime of its own; it holds the
mesh MCP server's account (the one sessions and agents call as), a person, the controller and an unproven caller as before.
Approval marks are for agent calls; inside the mesh a module may do everything. (novox, 2026-11-16 02:35 CET: the
indexer's runtime of its own refused its own `mesh-catalog.revision`, the very first view it waits on, so every reader
of the catalog waited for ever.)
- **A-394** (A-392; principle 3) A build seat's finding writer that cannot start is said in the module's journal and in
its health, and the module serves its steps on: a step's verdict never waits on a finding being written.
- **A-434** (A-433, §19; the operator, 2026-11-15 about 23:15 CET: "I only see 3 accounts, the mesh account, the agent
account and the operator account. The mesh accounts should have all the permissions"; "we can always add in gates and
guards later") Until the research on the mesh's accounts replaces it, every account the grants part composes but an
enrolling node's may publish and subscribe on every subject and put an ask on any seat; a seat verb answered by a
process and a tool runner's seat verb admit every caller; no seat loses its holder for every account hearing it.
Approval marks stay, for calls that are not a module's (A-435).
- **A-435** (A-433, A-434) Approval is the one mark kept: while a verb's seat definition is unread, a call of a verb
that may need approval from a caller that is not a module's account (the mesh MCP server's, a person's, an unproven
one) is held as before, never run unmarked; a module's account runs it.
- **A-436** (A-430b; principle 7) Until every node-engine reads a want by its version, a want for a node whose
node-engine reads parts carries the build's parts beside its version, and a module wanted with no passed pipeline
names the parts its `build.ended` made until its release is heard. Both forms go in the first change after G is
delivered, with A-430b's.
- **A-437** (A-428, §16; principle 7) The holder of `mesh-catalog` is wanted with no passed pipeline, as the pipeline
runner's own want is: every pipeline runs on a revision it took in, so it cannot wait on one. Every other module still
waits.
- **A-438** (A-426a; principle 7) For an acting wait, a running have whose node-engine states no health counts as
healthy, since an engine from before G states none; and a node that is online runs its node-engine, since online is
that engine's own word. Both go when every node-engine states health.
- **A-439** (principle 4) The assigner's tool `assigner_why` says why a want is what it is: the step of the pass that
decided it and what it read, kept as `why.<node>.<module>` in its state `written`. Until a manifest naming the tool
is taken in, `assigner_written` with `why` answers it; that stand-in goes with A-436's forms.
- **A-440** (the operator, 2026-11-15 about 18:35 CET) A build seat's queue serves its kinds in turn, so no kind of step
waits behind another's backlog.
- **A-441** (principles 3, 4 and 7; each a fault met live while G was brought up, fixed where it lives) These stand:
an ask's words cut a digest or fingerprint to 12 digits, so the operator channel does not withhold the ask as a
secret, and the backup ask names its composition by that prefix; the indexer writes the first registrations from the
module graph it holds, with no ask (the operator, 2026-11-14: a module in the module graph is a registration), and a
seed asked again writes what its key composed, ends no proposal key, logs each step with its time and prunes nothing
on a first registration (completing A-392); a pipeline step fetches a module directory only by its artifact's paths;
grants makes each seat's ask stream the bus lacks, from the seat list, with its part's account allowed to; a state is
read in pages, never a key at a time, and the indexer's tools keep their view of the trunk while its revision stays
the newest (a read over bus state); a get session's object reader lives as long as the session; the forge bridge
names each member of a group once and answers open groups or one group, never every ended one; grants reads its
loaded list one user per answer, so no answer outgrows one bus message.
- **A-442** (A-401; principle 7) The gate's start pass runs once; it asks again only while the registrations cannot be
read. A repository it leaves out is said and added at the next registration change or by hand, as A-401 says.
- **A-443** (principle 3) Each bus step of a distribution publish is bounded at 10 seconds and logged by name; a step
past its bound is given up and said, and its call is cancelled with it, never left running behind.
- **A-444** (the seat `git`'s contract; principle 7) The seat `git`'s `branch` with no branch named answers the
repository's default branch, as the forge names it; a build seat says each toolchain it reported, not only the ones
it could not.
- **A-443a** (A-443; principles 3 and 7) A step given up at its bound is not cancelled through the SDK, whose bus calls
take no context: its call ends at the SDK's own bound of the ask, and distribution logs, by the step's name, that it
gave the step up and, later, when the call behind it ended and how. No context is added to the SDK for it.
- **A-445** (A-379 replaces A-376 and A-378 for the start; principle 7) The bus keeper reads nothing of the seat
`mesh-catalog` before the first views: its A-376 path, and the manifest's consumes of `seat:mesh-catalog.views.modules`
and `.views.manifests` and invokes of `seat:mesh-catalog.modules` and `.manifests`, go, since the seat offers neither
those views nor those verbs and the validator refuses them.
- **A-446** (replaces A-195, A-206 and A-218, and §266's hold of pipelines for the runner's new version; the operator,
2026-11-16 about 10:35 CET: "yes remove the hold"; principle 7) The pipeline runner holds no pipeline for its own new
version: a take-in that changes the runner starts its pipelines as any take-in does, run by the runner that runs, and
the runner's new version is wanted once its own pipeline passed, as any module's. The state `bounds`' held keys, the
wait's `on` value `runner version`, the finding `pipeline-waits` and the verb `release` go. A pipeline a runner before
this one left held is started at the runner's start, as a newly started one. (Live, 2026-11-15 18:05 to 2026-11-16:
every pipeline was held for a new runner whose want waited on that runner's own held pipeline, so none ran a step.)
- **A-446a** (completes A-446; principles 3 and 7) When the runner places a pipeline whose steps were never asked, it says
`pipeline-started` for it, a pipeline a runner before it set aside included: the watchman ends a branch head's wait
on that event alone, and hearing it twice for one head ends the same wait again. What only the set-aside used goes
with it: the seat definition's `acts` entry for step `pipeline` and its 30-minute bound, the runner's wait `step` value
`pipeline` and the wait's `clock-from`, and the tools `runner_waits` and `runner_release`; G-5 §174 and §190 and G-10's
lines naming them are replaced by A-446.
- **A-447** (the build source of a Go artifact is its own module and what it loads, ADR 0267 §1; principle 7; the
operator, 2026-11-16 about 11:00 CET: "start the fix") The Go build module's fetch, asked with a module's directory
(`dir`), downloads, resolves and judges the go.mod and go.sum of that directory's Go module alone, and of each
directory its go.mod replaces with a local path, followed through their own replaces; never another Go module of
the fetched tree. A fetch asked with no `dir` (a check of a whole repository) does every Go module of the tree, as
before. The builder names the module's directory on every fetch it asks, not only for a module that names its paths;
the pipeline runner already does. (Live, 2026-11-16: every build and pipeline step of mesh-delivery's assigner,
builder, checker, teller, view and watchman at 901c800c failed on the indexer's `takein` go.mod, which none of them
loads; their fingerprints did not move with #219's fix, so they could never pass.)
- **A-447a** (completes A-447; principles 3 and 7) A module's own Go modules are the go.mod of its directory or, when
its directory holds none (a module at the repository's root whose Go code lies in `go/`, a contracts module with a
go.mod per seat), every go.mod under its directory (in a narrowed tree, under its paths); then each one's local
replaces, as A-447 says. A fetch asked with a `dir` that chooses no Go module fails, naming the directory: it never
goes on to steps that would pass with nothing tested or built.
- **A-447b** (replaces A-447's "a fetch asked with no `dir` does every Go module of the tree"; the operator, 2026-11-16
about 11:30 CET: "a build is also for a module, not/never a repo"; principles 3 and 7) A build, a step and a check are
always for one module. Every fetch of a build seat names its module's directory; a fetch with no `dir` is refused,
naming its asker, and no step runs on a whole repository. Every caller that asks a fetch with no `dir` today names the
module it builds or checks, in the same change.
- **A-447c** (principle 10, added 2026-11-16 at the operator's word: "nothing is ever repo-based") A-447b holds for
every part of building, checking and delivering, not only the fetch.
- **A-447d** (A-447b; principle 10) The checker's check of a pull request asks one fetch per module the pull request
changes, each naming that module's directory, and judges each module apart; it never fetches a repository whole.
- **A-447e** (replaces A-447d, and A-447b's refusal until then; G-6 §321; principles 7 and 10) The checker's check of a
pull request is removed with the checker at G's delivery (§321), so it is not rebuilt per module. Until it is, the
Go fetch's one caller with no `dir` is that check, and a fetch with no `dir` still fetches the tree whole for it;
every pipeline step and build names its module (A-447 to A-447c). The refusal of a fetch with no `dir` lands with
the checker's removal.
- **A-448** (§476, §493, §200; principles 3 and 7) `take-in.taken` names the catalog graph revision the take-in made
(`revision`), written by the indexer with the event. The runner starts each pipeline of a take-in at that revision,
never at the revision its own view holds; while its view holds an older one, it waits for it (§200) and places no step.
(Live, 2026-11-16: main.947 was started at revision 356 while its take-in made 357; its composition read 356 after
the indexer had removed it.)
- **A-449** (§153, §579, design 47's "kept while a delivery record pins it or it is current"; principle 7) A pipeline
pins its catalog graph revision as a delivery record does: the runner says the seat `mesh-delivery`'s `pinned` when it
starts a pipeline and `freed` when the pipeline ends, under the record id `pipeline.<key>`, so the indexer keeps the
revision while any pipeline runs on it. No new mechanism; who may say them is a grant of the runner's account.
- **A-449a** (completes A-449; principles 3 and 7) The indexer pins the revision a take-in made for each pipeline the
take-in starts, under `pipeline.<key>` (`<branch>.<source>.<module>` of each entry), in the same write as it keeps the
revision, so no later take-in can remove it before the runner hears of it. The runner's pin at a pipeline's start stays
(a set: saying it twice is harmless) and its `freed` at the pipeline's end removes it.
- **A-226b** (completes A-226 and A-167; ADR 0593 §7; principles 7 and 9) The seat `node-engine`'s `compose-candidate`
takes three key forms, parsed by one parser in the seat's client (`nodeengineclient`) that every caller builds its key
with: a check's candidate `<group>.<heads hash>` (as today), a branch pipeline's revision `<branch>.<revision>` and a
run's `run-<id>`. For a revision key the node-engine reads that catalog graph revision's entries through
`mesh-catalog.revision` and composes them as the candidate; for a run's key it reads the run's candidate from the
indexer's `runs` state. The seat definition's `candidate` input names the three forms. The assigner builds its key
with the client's builder, never by hand. (Live, 2026-11-16: main.948's composition step: "\"main.362\" is not a
composition key".)
- **A-226c** (corrects A-226b's run form; A-160; principle 7) For a run's key `run-<id>` the node-engine reads the run's
candidate from the indexer's `candidates` state under the group `run-<id>`, as it reads a check's candidate, with no
heads comparison; the `runs` state holds only a run's members and their ends.
- **A-450** (ADR 0663 §2, ADR 0621 §2; principle 7) A verb of the seat `mesh-catalog` that checks its callers takes the
node-engine's account `node.<machine>` as the module whose trunk entry serves the seat `node-engine`, and grants it
when that module declares the verb in `invokes`, as for any part. (Live, 2026-11-16: main.949's composition, "mesh-
catalog.revision is refused to \"node.novox\"".)
- **A-237b** (A-237, A-249; principles 3 and 7) The seat `node-engine`'s `compose-candidate` is a verb marked `chunks`:
asked with `input`, the node-engine sends the composition input in chunks (`buscalls.SeatChunks`) over its own bus
connection, whatever its size, and its answer names the chunks; the assigner takes it through its ordinary chunked
seat call (`CallFile`), as it takes the indexer's `input`. No size bound refuses an input any more. (Live,
2026-11-16: main.952's composition input was 1,000,641 bytes, over the 921,600 one answer carries.)
- **A-237c** (completes A-237b; G-6 §321; principle 7) An ask of `compose-candidate` that names no `call` token (the
checker's, until the checker is removed at G's delivery, G-6 §321) keeps the inline input and its 900 KiB bound; an
ask with a token gets the input in chunks only, with no bound. The assigner reads an inline input only from an engine
that answers no `chunks` (one before A-237b); both forms go with the checker.
- **A-451** (completes A-226, A-243 and A-252; principles 2 and 3) Where a machine's node-engine is not heard running
(no `running` of the module holding the seat `node-engine` on that machine, or one with no fingerprint), the
composition step builds that machine's composer from the revision's node-engine entry, the build the catalog says
the machine runs, and its log says so with the machine and the entry's fingerprint; the step does not stop on it.
(Live, 2026-11-16: main.953's composition could not run, "novox: no build of mesh-host is heard running there": a
node-engine running a build the catalog holds no entry for says no `running`.)
- **A-452** (extends A-405 and A-407 as ADR 0636 §3 does for claims; principles 10 and 3) A requirement nothing on the
node or in the rest of the mesh answers costs the modules that want it, not the machine: each such consumer stays on
the build it runs, its have saying why (the same "nothing provides" words, naming the provision), as a refused claim's
claimant does, and every other module of the node composes. The machine is refused only when what fails is the node's
own composition or no module could be composed at all. (Live, ace 2026-11-16 23:11: the node refused its whole
composition on ten unanswered requirements, eight of them of modules already staying on their builds with
`entry-missing` or no part digest, so the tool runner's released build 7df8e170 (`node-tools`) was never composed there.)
- **A-453** (ADR 0539 §16, §541; principles 3 and 1) The operator is asked about a backup only when a backup failed. Every
other hold the node-engine writes in a have (an entry asked or missing, §421; parts that differ; a kept data's form
not known) carries its `wait` with step `compose` and its reason, as §541 says; a hold for a failed backup carries
none. The assigner raises the backup ask of ADR 0539 §16 only for a held have with no `wait`, and withdraws a backup
ask whose have now carries one or is no longer held. (Live, 2026-11-17 00:32: the assigner asked "The backup of avahi
failed on ace" and the like for about forty modules held only for `entry-missing` or `entry-asked`.)
- **A-453a** (A-453, ADR 0539 §16; principle 3) "No longer held" in A-453 means the have has moved past the hold: running,
removed, or at another fingerprint. A have in the middle of a move (`backing-up`, `applying`, `checking`) is not "no
longer held", so the backup ask stands while the node backs up again after an "again" answer, and a second failure
is asked under the next number, as §16 says. "At another fingerprint" applies only to a have that is not held: a held
have names the build it keeps running, which on a failed backup's hold is never the want's, so a held have has not
moved past its hold (corrected 2026-11-17, from the review of A-453's assigner half).
- **A-454** (§421's parts check, A-432; principles 7 and 3) A want's parts are the artifacts the mesh built for the
module. The node-engine's check that what a composition fetches matches the want's parts counts only what is fetched
from the mesh's artifact store (a bundle, an archive, a program, or a container image the mesh built, named through the
artifact store or the registry seat); a container image the manifest pins from outside (an upstream image by its digest)
is no part and is never "fetched and not in the want's parts". (Live, novox 2026-11-17 00:46: distribution 0.1.11 held
"parts-differ: fetched and not in the want's parts" for the upstream `registry` image its manifest pins.)
- **A-455** (§369, §588; principles 4 and 3) A trunk build that starts while its series is reserved by an older build of
the module that was never released takes the reservation over: the newer build reserves the series' next version, the
older one, which no want follows any more, is dropped and said so in the builder's log. And a done trunk build that
reserved no version is given one by the builder's tool `builder_reserve_done` (module, fingerprint, why; a hand act):
it reserves the next version of the series for that fingerprint with the build's artifact digests and queues its
publish as §588's publish-again does. (Live, 2026-11-17 00:49: the tool runner's (`node-tools`) 0.1.13 stayed reserved for build b5c7a048,
whose entry was gone, so the passed build 7df8e170 of main.984 reserved nothing, was never released, and its wants
carry no parts; a done key cannot be built again.)
- **A-454a** (A-454; principle 3) The A-454 exception applies only to the direction "fetched and not in the want's parts":
a digest the want's parts name always counts as fetched when the composition fetches it, whatever marks its
resource carries; a fetched digest the parts do not name is counted only when it is the mesh's own (named through
the artifact store or the registry seat, and not an upstream copy). (From the review of A-454: a mesh-built image
whose seat holder is not on record carries no `image-seat`, so skipping it would hold the module on its own part.)
- **A-456** (§369, §588, A-455; principle 3) A pipeline's build step passes only when the build it judges has a
reserved version for its fingerprint. A build that ended done with no reservation fails the step loudly ("built at
<fp>, but no version is reserved for it"), naming `builder_reserve_done` as the recovery, so no pipeline says
"passed" for a build that can never be released. (Survey 2026-11-17 00:58: the tool runner's build 7df8e170 passed
main.984 and was never released.)
- **A-457** (§588; principles 1, 3 and 4) A version's publish is work the builder owns until `version-released` is
written for it: every reservation of a done build is a publish item on the builder's own work queue, asked when the
build ends, redelivered after a failed or unanswered publish with a growing pause (a setting, principle 11), and
acknowledged only once the release is written. At its start the builder puts on the queue every reservation whose
build is done and that has no release, so what stood before this change moves too. A publish that fails its
setting's number of tries is said loudly (log and condition) and stays readable in the builder's tools; it is never
waited on a `running` or `manifest-reads-changed` event any more. (Survey 2026-11-17: 115 reservations, 108 built,
3 released.)
- **A-458** (§354's composition step; principle 3) The composition step judges by what the composer said: when it judged
no candidate (every verdict "not judged"), the step ends "could not run" with the composer's words, never "passed";
a step passes only when at least the candidates it was asked are judged and none refused. The asker (the assigner)
names the seat list in the request's `seats` from the seat list it holds (ADR 0636), so the composer has what it needs
to judge. (Survey 2026-11-17: 86 of 141 pipelines "composition could not run", and the code read "not judged" as a
pass.)
- **A-459** (the assigner's rule 7; principles 10 and 3) Rule 7's "composed" is asked per module: a module waits only
on the composition of its own have at the fingerprint it waits for, on that node, never on whether the node composed
anything else. One module that does not compose on a node never holds another module there. (Survey 2026-11-17:
`composed(node, module, fp)` read the node's own have and ignored module and fingerprint, so one node that composes
nothing held every module on it.)
- **A-460** (A-452, ADR 0539; principles 10, 3 and 2) A node-engine's install of a build of itself that fails (the
version's folder cannot be made, its archive not fetched) costs the node-engine's own have only: it stays on its
current build, its have says why with the error's words, and every other module of the node composes. The
node-engine makes its versions folder where its own unit lets it write; a refusal there ("operation not permitted")
is a fault of the node-engine's install and is fixed in it, not on the machine. (Live, ace 2026-11-16 23:33: the
node composed nothing after `mkdir .../versions/88216c0e9496.unpacking: operation not permitted`; its want for the
node-engine names an older build, d9e5d175, because no pipeline of the node-engine passed at a newer one; that want
moves with this change's own passed pipeline, by §29.)
- **A-461** (A-452, A-407; principles 2 and 10) Where a module on this node needs another node's module (a store
address, a provider's endpoint) and that module runs a build that has no entry, the node lays it from the entry of
the build its want names, which is the mesh's truth of what it runs there. Where neither build has an entry, the
holder stays name-only and only the modules that need its address wait, as A-452's consumers do, with the same
words; the node is never refused whole for it. (Live, ace 2026-11-17: ace laid the artifact store's holder
`distribution` name-only, novox running 44a32f00 with no entry while its want names 4133e115, which has one, and
refused its whole composition, "the artifact store has no address".)
- **A-462** (§29, A-428; principles 2, 5 and 3) A want of a module that names no version and no parts while the build
it names has a released version is completed by the assigner in one pass of its own: the want keeps its
fingerprint (nothing moves, so §29's passed pipeline is not needed for it) and gets that version and its parts. The
pass runs at the assigner's start and on every `version-released`, and says per want what it filled. A want whose
build has no entry is not filled: the indexer is asked to take that build in again (`indexer_take_in_pending`), and
the want says it waits for the entry. (Survey 2026-11-17: 40 to 80 modules per machine on wants from before G with
no version or parts, `entry-missing` or "want names no digest", and a want was rewritten only after a passed
pipeline, which waited on A-458.)
- **A-459a** (A-459; principles 1, 10 and 3) Rule 7 learns "composed" per module from the node-engine's `composed`
for `<node>.<module>.<fp>` (its digest a composition, a refusal not composed), followed per node as the assigner
follows the haves, through the read its manifest already declares (`reads`: `mesh-host.composed`); the module's own
have is not the source, since none names a fingerprint before its want does. (From the develop of A-459:
`composed(node, module, fp)` read only `<node>.node`. Corrected 2026-11-17: the first wording said "never by
reading the node-engine's bucket", which contradicted the declared read the haves already use.)
- **A-462a** (A-462; principles 7, 9 and 3) "A build that has no entry" is a fingerprint the catalog half the assigner
follows holds no entry for on any watched branch. The assigner asks the indexer nothing for it: taking a build in is
the indexer's own work on its pending state, and its tool `indexer_take_in_pending` takes the indexer's own keys,
not a module or a fingerprint. The want says, in its words, that it waits for the entry of that fingerprint, so the
wait is loud and read with `assigner_why`; that sentence of A-462 about the indexer's tool is withdrawn. (From the
develop of A-462.)
- **A-456a** (A-456; principles 3 and 7) A released build counts as reserved: the build step passes a reserving build
whose fingerprint has a reservation or a released version. The builder knows the second from its own record of the
`version-released` it wrote for that fingerprint; where that record keeps no fingerprint today, the builder keeps
it there (its own state), never asking the catalog. (From the develop of A-456: the reservation is deleted at
release, so a rerun of a released build's step failed "no version reserved".)
- **A-457a** (A-457; principles 7, 1 and 11) "The builder's own work queue" is no new stream and no new process: the
publish item is a re-ask in the builder's own re-ask state, the mechanism `builder_reserve_done` and publish-again
already queue on, consumed by the builder's one process. The value carries its tries and its next try; a failed or
unanswered publish is written again with tries plus one and a pause of `publish-pause` doubled per try up to
`publish-pause-max`, and after `publish-tries` tries it ends loudly (log, condition) and stays readable in the
builder's tools; all three are settings with defaults (30 s, 30 min, 8). (From the develop of A-457: a work-queue
stream would need a `bus` restating the builder's whole account, or a second process.)
- **A-461a** (A-461; principles 7 and 1) "The build its want names" for another node's module is read from what the
composer already holds for every node: the assignment's placed build of that module on that node (absent meaning
the newest delivered, as the assignment says). The composer watches no other node's wants. (From the develop of
A-461: the composer holds only its own node's wants.)
- **A-460a** (A-460; principles 2, 3 and 7) The versions folder's refusal on ace was a hand act (an immutable mark set
2026-11-13, lifted since), not a fault of the node-engine, so A-460 needs no change to its install. A have that
failed to install is not tried again on every pass: it is tried again when its want changes (a new fingerprint or
version), which is how it leaves a stale failure; the node-engine's want on ace still names an older build, and
moves when this change's own pipeline passes (§29). Trying the failed install of that older build again would put
an older node-engine on the machine, so no retry is added. (From the develop of A-460.)
- **A-461b** (A-461a; principles 7 and 9) Where the placement names no build ("the newest delivered"), the composer
takes that module's trunk build from the seat `mesh-catalog`'s verb `module`, as it already does for a module an
assignment adds (`addedModule`), and lays the module from that entry. A-461's fallback applies only to another
node's module, and only when its running build has no entry; and only a store with no known address (the holder
opens no listen, or two) holds the modules that fetch from it: every other fault of finding the store still refuses
the node, as before (principle 3). (From the review of A-461: novox's placement of `distribution` names no build.)
- **A-456b** (A-456a, A-457a; principles 7, 11 and 3) The builder's record of releases is one key per module,
`released.<module>`, holding its last `released-kept` releases (fingerprint and version; a setting, default 16), the
oldest dropped when a newer one is written; a rerun of an older build's step than those is judged as a build with
no reservation. A publish item is deleted when its reservation is dropped or taken over, and when its release is
written; never because its artifacts are published, so a release that never comes is tried again. At its start the
builder asks an item for every reservation of a done build, whatever its artifacts. (From the review of A-456a and
A-457a: one key per release filled the 1 MiB state for good, and an item went before its release.)
- **A-457b** (A-457, A-457a; principle 7) The builder's setting `republish-max-versions` goes: it bounded what one
`running` or `manifest-reads-changed` event queued for publish, and A-457 retired that path, so nothing reads it.
A value set for it in the mesh's configuration is left unread and may be removed there. (From the review of A-457a.)
- **A-459b** (withdraws A-459 and A-459a; ADR 0571 §6; principles 7 and 10) Rule 7 reads the node's own composition,
as on main: a node names its composition only after it composed what its wants name, so no node holds a module's
composition at a fingerprint before a want names it, and a per-module, per-fingerprint read would hold every new
build on every placed node. That one module costs only itself is the node-engine's part (A-452, A-461): a node that
composes the rest names its composition, so no module's fault holds the others. (From the review of A-459: four
order-wait tests fail and no placed node moves once the fixtures write compositions as the node-engine does.)
- **A-462b** (A-462a; principle 3) A want a rule holds keeps that rule's words, which say why it does not move; A-462a's
words ("waits for the entry of that fingerprint") are written only on a want no rule holds. (From the review of
A-462.)
- **A-463** (ADR 0573 §2; principles 3, 1 and 4) A merge the indexer never took in is noticed and taken in by the
indexer itself: at its start and on every source revision the gate's source names, it compares each repository's
`taken` merge with the head that source names, and a repository whose head is newer is taken in as its merged
message would have been (the catch-up of `indexer_take_in`, for that repository), said in its log, with a condition
raised while it stays behind (the repository, its taken merge and the head) and cleared once taken. A merge is never
left untaken in silence, so no package stays published at a commit older than its trunk. (Live, 2026-11-17 02:00:
mesh-contracts #14, merged 2026-11-16 14:14Z, was never taken in; `packages` kept every mesh-contracts package at
20b06dbc, and every build of the node-engine since 15:24Z and of the assigner since 22:51Z failed on symbols only #14
holds.)
- **A-464** (amends A-463; principles 2, 3, 1 and 4) The gate's source of a watched branch follows every merge into
that branch, whoever made it: on the forge bridge's `merged` of a repository the source names, at the branch the
source watches, the gate moves that repository's head to the merge and says `source-moved`, as for a group it
merged itself; a merge it did not make is said so in its log and its record. At its start the gate compares each
head with the forge's branch head (through the seat `git`) and moves every one behind it the same way, so a merge
whose message was lost is followed too. The gate's tool `gate_source_follow` (branch, repository, why; a hand act)
does the same for one repository now. A-463's comparison in the indexer stays: it then takes in what the source
names. (Live, 2026-11-17 01:45: the source still named mesh-contracts at 20b06dbc, the node-engine's `mesh-host` at 13fcfae0 and
mesh-delivery at 8c4489fb, though #14, #232, #233 and #237 were merged into their mains; every merge made outside
the gate, as in recovery, was never taken in, built or delivered.)
- **A-464a** (A-463, A-464; principle 4) The indexer's state `takeins` is read with the tool `indexer_read_takeins`, as
`indexer_read_taken` reads `taken`: every key, or one key, with its ask, whether and when the indexer answered it and
its answer, and how full the state is against its cap. A take-in asked by hand and never answered is then seen, not
guessed. (Live, 2026-11-17 01:39: the catch-up of `indexer_take_in` answered "in takeins.catch-up", and no tool could
read that key.)
- **A-464b** (A-464; principles 7 and 3) A merge the gate did not make is one whose `merged` names no delivery group of
the gate. "Said so in its log and its record" is its log line (the repository, the branch, the merge and "not made by a
delivery group of the gate") and its `source-moved`, which then names no group; the gate writes no other record for
it. (From the develop of A-464: the gate holds no merge state of its own for such a merge.)
- **A-464c** (A-464; principles 2 and 7) The gate moves a head of its source only forward and only when it changes: on a
`merged`, at its start and through `gate_source_follow`, it moves a repository's head to a merge only when the head it
holds is contained in that merge (through the seat `git`), and it writes no source revision and says no
`source-moved` when the head would stay the same, as on `take-in.taken` for a head a `merged` already moved. A merged
message heard again, late or out of order is so left without effect, said in its log. (From the develop of A-464: a
replayed message moved heads backward, and each merge wrote two revisions.)
- **A-463a** (A-463, A-464; ADR 0633 §4; principles 3 and 7) A head the source names that the indexer's pending holds
(a merged message waiting for its group's end or its build) is not behind: it is taken in by that path, whole with its
group, and raises no condition. The comparison takes in, through the catch-up, only a head that no pending item holds
and whose taken merge it contains. A source revision heard while a catch-up runs is compared again when that catch-up
ends, so none is skipped. (From the review of A-463: under A-464 the source moves at each member's `merged`, before its
group ends, so a catch-up there took one member of a group in alone.)
- **A-464d** (A-464c; principle 2) Every move of a head of the gate's source is forward only, the move at `take-in.taken`
too: a head moves only to a merge that contains the held head, and a later merge that already moved it is never moved
back. (From the review of A-464c.)
@@ -0,0 +1,152 @@
---
layer: to-be
status: in-progress
design: 47
milestone: G
updated: 2026-11-16
decisions:
- "ADR 0667"
---
# 47, milestone G — what is owed once G is delivered
The list of what waits for G's delivery: the operator's remarks deferred while G was brought up, the stand-ins the
lead's answers mark to go after G, the hand state of G's live start, and the faults met on the way. It is checked when
G is delivered: each item is then removed, taken into a design or a record through its playbook, or noted on the issue
that covers it (asking the operator before a new issue). An item is struck from this list in the same pull request that
deals with it, which is how the list is kept true.
## The operator's remarks (2026-11-14, deferred until G is live)
1. **Two graphs only:** the catalog graph and the assignment graph. Each node's modules publish the state they own and
the version they run on the bus; a gap between the assignments and what a node says is closed by the node-engine.
2. **What a node can run is checked only on assignment and un-assignment** (for the modules that depend on it), never
as an input to every computation. The glossary's *assignment graph* and ADR 0656 say the assigner computes from the
catalog graph, the placements and the nodes' capabilities: check the record and the code against this.
3. **Assignments enter as commands** (the seat `mesh-module-assigner`'s `propose`: place, take off, ...); the assigner
computes no placement, it only derives each placed module's version from the catalog graph into the wants. Check the
glossary's wording against this.
4. **A pin** ("use this provider for this provision") may not belong in the assignments at all: question it, and what
the provision wiring should be instead.
5. **Who says a node's running versions:** the node-engine (its haves) or each module itself. The operator's picture
is each module; check which the design means.
6. **One address for a seat's view.** A reader's manifest names the assigner's view by its seat, the reader's code asks
the holder module's tool by its module name, and the grants part translates one into the other (and gave only the
events, not the right to ask: the indexer's first view was refused at G's live start). One address, the same in
manifest, code and rights.
7. **A registration is part of the module graph, not a list beside it.** After G's start the module graph held 178
modules while the registrations held none, with a seed ask to copy them from the old registry. Check the design
against one graph whose entries are the registrations and what each registered repository says.
8. **One bookmark (durable consumer) per process, never shared** (A-397). A module's program and its tool runner read
events under one durable name; the runner took all 178 first registrations and the program heard none, silently.
Share a durable name only for shared work (a queue). Owed: amend design 19's "one durable consumer per module"
(written when a module was one process) and the lines of designs 25, 32 §3, 57, G-5 and G-8 that follow it; build
per-process names in the grants part, mesh-sdk's process hub and the tool runner, with a start refusal and a log
line; the test of two processes of one module both hearing one event on a real bus; a check that a durable has one
reader; and a design-review playbook question: "what if two processes of one module do this?"
9. **The same mistake elsewhere.** A survey at G's start found it live in three modules (the assigner, the indexer and
the grants part: the runner acks and drops what no tool view wants) and possible in every module whose tool runner
arrives later (builder, checker, forge bridge, pipeline runner, teller, view, watchman, distribution, recovery).
Already avoided: the bus and zfs keepers (A-340), the node-engine, work queues. Also: a seat verb served by both a
program and its runner in one queue group splits the answers between them; durables nothing reads keep pending
messages.
10. **`registration-changed` is said for a first registration too:** check the event's name against what it says.
11. **Silent waits met during G's start:** the indexer's first versions fill returns without a word while another fill
runs, so a hung fill held every later one silently for two hours; `grants_membership` and `grants_assert` time out
after 30 seconds with no log line; the tool runner says a dropped event once per key.
12. **The module graph's edges are dependencies, never verbs or bus subjects** (2026-11-16; principle 9): a manifest
says what its module depends on, a seat where the need is meant to be interchangeable and otherwise a module (a
build dependency on a package module, a program needing another module), and the graph's edges come from that
alone. The verb edges of ADR 0614 §2 and §3, and the cycle exceptions they needed (A-361, A-428, A-437), go; whether
`invokes`, `uses` and the `views` forms stay in a manifest at all is asked with the research on the mesh's
accounts, and "each fact is declared in one place".
## Met during G's start (2026-11-14)
- **The seed of registrations** left out modules whose manifest lives in the application's own repository on its
watched branch, so they were assigned but unheld and the assigner refused every proposal (ADR 0567 §2). The seed
should register every assigned module from where its manifest is.
- **The indexer's catch-up and take-in** say nothing in the indexer program's journal.
- **The provisioner** drops a published composition with neither plan nor refusal silently and then removes the
node's plans; no read tool shows a node's composition.
- **The node-engine** is refused the builder's `build.ended` by the bus every five minutes: the grants part does not
give it; to fix before builds matter.
- **The register verb refuses an unprotected watched branch:** the operator approved removing the check (from the
seat `mesh-catalog`'s definition and the indexer that serves `register`).
## Stand-ins the lead's answers mark to go after G (2026-11-16)
- **A-436 with A-430b:** a want carries parts beside its version for a node-engine that reads parts, and a module
wanted with no passed pipeline names the parts its `build.ended` made until its release is heard.
- **A-438:** for an acting wait, a running have whose node-engine states no health counts as healthy, and an online
node runs its node-engine; goes when every node-engine states health.
- **A-439:** `assigner_written` with `why` answering for `assigner_why`.
- **A-434:** every account may do everything, and every seat verb admits every caller, until the research on the
mesh's accounts replaces it (the operator, 2026-11-15: three accounts, mesh, agent and operator; approval only as a
per-tool mark for agent calls).
## Hand state of G's live start (2026-11-16)
- **Runtime drop-ins on the control-node** that pin a hand-made fingerprint (and the tool runner's environment) for
the grants part, the teller, the forge bridge and the tool runner. A delivery does not remove them: each goes before
its module restarts on the delivered build, or the module believes it runs the old one.
- **Hand-placed builds** (the indexer and its tools, the tool runner, the control-node's node-engine, the assigner) are
replaced only by their deliveries; check each is.
- **About 80 hand-built modules on the control-node have no catalog entry:** their first move is held for a backup
(the kept-form hold).
## Faults met on the way (2026-11-16; note each on the issue that covers it)
- `node-files.delete` cannot move a root-owned file to the trash; `node-service-manager` has no daemon-reload.
- The indexer does not stop on SIGTERM: it keeps draining build-log entries until systemd kills it.
- The backup ask's headline says "backup failed" for a kept-form hold.
- The assigner logs a line naming about 140 modules every one to three seconds, and did not raise the backup ask
again after a restart.
- `mesh_wait` refuses the operator channel's events: `operator-channel` is neither a seat nor a module of the catalog
half.
- Four tests of mesh-delivery fail on main, reading files that do not exist there: `sdk`'s
`TestThePackageArtifactNamesNoPathForTheWholeDirectory`, `sdk/buildstep`'s two fixture tests and
`meshdeliveryclient`'s `TestTheRemovedVerbsAreTheDefinitionsRemovedRows`.
- The node-engine's repository holds about 1721 unit tests against the budget of 500.
- The live indexer was seen refusing the assigner's manifest for claiming `mesh-placement`, which neither main nor
G's branch claims: confirm the verdict after G's take-in.
## Tests run against what runs (the operator, 2026-11-16 about 13:00 CET)
- **A test runs against the bus as the mesh composes it.** The pipeline runner's integration test made its own
shared event stream catching every seat event, so a step's end reached the runner there; on the live bus a step's
end is a kept event on a stream of its own, which the runner never bound, and no pipeline could pass (fixed in the
runner at G's start, 2026-11-16). Wanted: a principle ("a test runs against what the module runs against: the bus as
the mesh composes it from the manifest, never a layout written in the test"); the shared library gives an
integration test the streams, kept streams and consumers the composition makes for the module's manifest; every
integration test that creates its own streams moves onto it (count them first); the runner's test publishes a step's
end on the kept stream.
- **A handler of a kept event is heard or fails loudly.** Starting a program's events binds only the shared stream;
kept seat events need a second call, and a program that registers a handler for one and never makes it hears
nothing and says nothing. The shared library binds the kept streams of every registered handler itself, or refuses
to start.
## One way of asking work (2026-11-16)
- The builder, the pipeline runner, the indexer (its validation before a take-in) and the checker each ask the build
seats in a shape of their own; the validation rides the check queue behind pipeline steps (a 6-second validation
waited over 20 minutes). Parked by the operator for research right after G: one generic pipeline for work and gates.
- An ask given again after a fix keeps its old form (a step asked with no module directory came back without one).
- The builder asks a failing build again every two to three minutes, reserving a new version each time.
- The builder's queue tool lists only the oldest 50 waiting entries (left over from 2026-11-14), so it cannot show
what waits now.
- The pipeline runner serves no tools on the control-node and logs nothing per step; the build seat's step events say
no pipeline before a step ends; `mesh_wait` refuses the build seats' and the runner's events; the bus's stream reader
cuts a body at 4 KiB; a running test step logs nothing.
- `TestTheBuildModulesKeepNoCopyOfTheDeliverySDKsStepCode` fails on mesh-catalog's main (the empty
`modules/mesh-build-go/buildstep` directory); the seven `checks/` tests, nats-tools and distribution fail in a run over
the whole of mesh-catalog.
- The glossary's **pipeline** and **pipeline step** entries leave out the build, which the runner asks as the step
`build` of the seat `mesh-builds`: correct them.
- A pipeline the runner set aside before A-446 with steps already written (`main.940.mesh-delivery-sdk`) is started
again by neither of the runner's start paths.
- The node-engine's journal and the have's phases say a module is "held" and that the node "keeps running what it
runs" for one thing: a module left on the build it runs, its new build not applied. The operator, 2026-11-17, wants
those words gone: retire "held" and "kept" in that sense from the node-engine's lines and the glossary, with one
glossary word in their place.
- A composition that passes on every machine as "not judged: the composer was handed no seat list" is reported as
passed: a composition step that judged nothing must fail or say so (principle 3).