A node's plans took 2.6-13.6 s of CPU on the provisioner: one composition
resolved every machine of the mesh some two hundred times (who is on the
private network, and the rest of the mesh, worked out again on every
look), three compositions did so again each, and the holder composed
twice. A composer now works each out once over its unchanging Input;
Composing gives Resolve, Render and Generators one composer;
PlansHeldTwice composes the wanted and the newest builds over one read of
the bus, says where the time went by phase, and the states reader keeps a
build's take-in by its content. WatchedStates takes a process's own
writes into the copy at once, so a holder reads its own states from it.
Measured on a captured novox mesh: 11.7 s to 0.49 s per composition of an
unchanged node, plans and provisions identical on all four nodes.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
The engine composes its node's wants itself, so the derivation the controller keeps
privately moves beside the layout: DeclaredPackagesOf gives the mirror's bytes from the
compositions, DiffDeclaredPackages what the node seat's verb packages compares until the
lease write, and EnginePackagesRead names the state as readers declare it (mesh-host.packages,
as mesh-host.data). Each engine may read the controller mirror's key of its own node alone.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
mesh/check pass 67354f8e019a on cad8250aaa8c | unit 7056 run, 243 kept, 16 skipped | integration 7194 run, 275 kept, 25 skipped | over bound: unit, in…
A check ask named no toolchain, so a node on the minority version could run it and its pass be recorded as run on the agreed one. CheckJob now carries the toolchain; a node on another declines it before it fetches, as for a build ask.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
With the toolchain agreed by majority, a node on the minority version can pull an ask for the agreed one. It now declines it (the issue 890 path) for the asker to put again, so it never fails a build.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
One build node on another version stopped every build of the language. The agreed toolchain is now the strict majority's; a tie still waits and names the versions.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
mesh/check pass 790d1c82d027 on da41823855a9 | unit 1964 run, 5697 kept, 30 skipped | integration 1999 run, 5829 kept, 38 skipped | over bound: unit, …
A read could start a state's watch before its follower joined, and a change landing in that gap woke nothing (review
of ADR 0580 §2): a joining follower now runs over the copy's current values before any later change. The watched
reader keeps each archive once per URL, and both caches are bounded.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
The controller's hand-act and warranted are gone (issue 1953 L5b) and the bus refuses them, so the recorder no longer
falls back to them: an act the seat does not take is an error saying nothing was recorded. And ADR 0580 clause 1's
reader is shown reading a build by key each time and a manifest once per digest.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
A whole state read key by key took the provisioner's composition of one node past 30 s and the assigner's compare past
its call bound (ADR 0580 clause 1): each state is now one watch's current values, kept by its changes; a build is read
by key and a manifest once per digest.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
mesh/check failed: unit: go.gofmt: not as gofmt writes them (run gofmt -w): go/catalogue/streams.go | 62cddf4efc62 on ff5e00f1f894 | unit 0 run, 78 ke…
Review round 1 (issue 1585): L1, DataItemsHeld holds a module by the culprit rule PlansHeld uses, a held module's
data left out, so an item its composition refuses holds that module alone; L2, mesh-sdk's own tests scan its go/
tree as mesh-catalog's checks/sdkjudge does, with the same list of composition's refusals. Two tests merged into one.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
ADR 0573 §2 has every module build against the latest published version of
each mesh package it uses, chosen by the planner and written by the build,
never pinned by hand (issue 1585, S4). The ask, the fetch request and the
merge preview now carry the packages and their versions, and the new-form
fingerprint takes each version, so a package's trunk merge changes every
dependent's fingerprint. A module using no package keeps its fingerprint.
Folded onto feat/judge-once by builder-1585a: Module keeps both Packages and the verdict fields.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
Issue 1585, S19: the strict parser, every manifest problem function and the judge's tables (192 declarations,
about 5,200 lines) and their 244 tests leave; composition keeps its own refusals and the helpers it uses. The
catalogue's tests decode the manifests they build. ParseManifest, ParseManifestWith, ParseManifestForInputs,
CatalogueProblems and the rest are gone; mesh-delivery-indexer/judge holds them.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
Issue 1585, S6-S8: ReadTreeManifestWritten and WithTemplatesFrom decode (the judge in mesh-delivery judges what they
answer); Manifest.Predefined, SetPredefined and FormsWritten expose what the moved judge reads; buscompose names its
removals without the catalogue's allow-list; tests outside the catalogue decode, two tree-judging tests moved.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
Issue 1585, S11 (mesh-sdk half): Case.Verdicts carries the indexer's verdicts; a refused module is held as its
stub and named; an unknown field is set aside and named, never refused.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
Issue 1585, S2: one refused manifest silenced disk-load for 45 hours. frombus decodes leniently, takes the
indexer's verdict, stands a refused module in as its stub and names it in Input.Refused; DataItemsHeld names it.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
Issue 1585, S3: state mesh-delivery-indexer_verdicts keyed by manifest digest, event manifest-refused, a modules
value carrying its verdict, and TakeManifestVerdict reading a missing verdict as accepted. Contract only, no rules.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
mesh/check pass c28c645ca406 on 5ae69df8ccde | unit 4510 run, 4259 kept, 49 skipped | integration 4589 run, 4361 kept, 51 skipped | over bound: unit, …
The contract (state name, key form, value) is in delivery; buscompose grants the engine its own
node's keys of mesh-host_data and each node's disk-load a read of its own node's keys alone, the
same way the engine's other states are granted.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
mesh/check pass cbbfbf3d2d55 on fea48f19cea5 | unit 2729 run, 4570 kept, 16 skipped | integration 2975 run, 4493 kept, 25 skipped | over bound: unit, …
A finding's field `token` is named like a credential, so the tool runner
refused every finding that carried it; it is `ref` now, and a finding
written under the old name is still read. A module the tool runner
carries got no grant for its own held seat's events, so the self-check
could not say `checked`; a held plain seat's events are now in the
holder's seat traffic, its membership and its runtime's grants.
Quick vs thorough: quick, these fixes for L4's take-over now; the
thorough fix (research 032 / ADR 0240's health view extension) is
decided later.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
mesh/check failed: integration: mesh-vault/cmd/mesh-vault.TestTheVaultWholeMintsOnTheBusWhoeverHoldsTheLease: never: the account minted under the cont…
influxdb and unifi publish bare ports that mesh-placement assigns none: the machine runs them at
the software's own number, so a plan refusing them held their whole plan for a route that works.
route-proxy served every route at the software's own port (gitea's 3000, published at 20000;
grafana's 3000 on another machine, published at 20012) and public sites answered 502. The plan's
contribution now carries the port the placement assigned, the number ${port:N} gives the module,
and a composition for plans refuses a port none assigned by module, node and port.
mesh/check failed: shanks: the check of novox/mesh-controller#487 failed: go mod download in modules/route-proxy: go: git.novox.be/novox/mesh-controll…
The node key <node>.node is a write's commit: it names each placed module with its value's digest, the
module keys carry the value they replace until it lands, so a reader never composes half a write. Every
reader goes through PlacementKeys/PlacementsOf, which read the legacy placement.<node> too while the
change is delivered.
mesh/check checking 748543c35f7f on 712ec9481b40 | novox/mesh-sdk#147 in group fix/engine-seat-any-verb
The grants part composed the engine's subscriptions from catalogue.EngineVerbs at its own mesh-sdk pin, so mesh-host #184's optional verb composition was refused on every node and every engine reopened its link every four minutes. Quick fix: re-pin the grants part. Thorough fix (this): the seat's subjects on the engine's node, any verb, as a module answers mesh.mod.<module>.tool.>; callers stay granted per verb.
mesh/check checking 8ed3b872e0ba on dec3bfe7a8cb | novox/mesh-sdk#145 in group fix/held-keeps-contributions
A module held for its own secret keeps its contributions (leaveOwn). When its grant to the provider was not issued
yet, the provider's grants file carried "foreseen: made by the next send" for it, and the applying path passed over
stand-ins of a module already held. Such a stand-in now leaves that provision's contribution out of the held module.
A provider waiting for a yes said why only in its have's words; nobody could read the composition that waits beside
what runs. The engine's read verb composition answers that, and ChangedResources lists what a move changes, a
received file naming the consumers that join or leave.
A module held for an own secret never made (nextcloud's admin on novox, 2026-10-30) was composed as a bare stub, so
its contributions left postgres's grants and keycloak's clients, and their helpers would have retired its login.
The culprit now tries leaving out its own secrets alone before its whole offer.
JudgeMove said a changed received-, bound- or secret- file as "where its data is kept"; it now names the consumers
that join or leave. The judgement stays a move: its yes is what kept nextcloud's login.
Quick fix: leave out own secrets alone before the whole offer. Thorough fix (not here): carry a held module's last
contributions from what runs into its providers' received files whatever part of it fails to compose.
jschoubben
mesh-tools