Registry milestone: mesh-catalog (issue 1966) #612

2026-11-04T17:50:44Z

mesh-tools commented

2026-11-04 17:50:44 +00:00

The registry milestone (issue 1966; ADR 0605-0636 as hq's designs name them), taken in onto its trunks after milestone 47/B0 as milestone 47/G's first step (design 47; issue 1921 notes 54-61).

One delivery group, branch feat/registry-milestone in mesh-sdk, mesh-contracts, mesh-delivery, mesh-host, mesh-tools, mesh-catalog and mesh-networking: one check, one Approve. No member names another's version (ADR 0573); every mesh package is required at the zero version.

This member: mesh-build-go and mesh-knowledge take the trunk: one invokes list (the automatic merge gave it twice), the run step on the delivery SDK's buildstep types and the checker's client, the manifest asked of its owner.

The registry's reviewed heads (three rounds, 2026-11-04 to 11-06) were never pushed until now; the check of 068b7d23cf1a judged the heads of 2026-11-04. The trunk step was prepared and reviewed in three rounds (2026-11-07): the merge resolutions keep B0's facts and checking-program rules and the registry's generated types; every check composes with the composer at the node-engine member's head, else its trunk; a module the running indexer refuses but the check's own validator accepts is composed and named in the verdict with its build awaited, never passed silently; the checking program answers only new or changed seat definitions when given the reader's digests, and without them (the running B0 checker) only each seat's own description, 171 KB under the 256 KiB bound.

Rules bent, said here: mesh-delivery stays over its test budget (662 unit, 68 integration; was 675 and 68) and mesh-host over unit (1745; was 1746), falling. Out of this step, noted on issue 1921: three asks carry the whole seat list (~430 KB) bounded only by the bus's 1 MiB; the trunk case of the manifest validator is given the heads' seats.

After check 616c42434ac2 (issue 1921 notes 63-64), reviewed again: mesh-catalog requires the module of meshdeliverycheckerclient, and a new check checks/requiredmodules refuses an import whose module the go.mod does not require; mesh-delivery's assigner test flushes after its subscriptions (a race), and the checking program's seat answer carries no verb or message schemas (the tool runner's state refuses a schema property named credentials).

After check 8a80e8e86858 (issue 1921 notes 66-67), reviewed again: mesh-catalog ignores the new check's program and mesh-build-go's test workspace lays what the composer requires; mesh-host composes from the change's manifest wherever a node's build was made of another (by its manifest digest, or by its contract where the running node-engine names none), naming each such module as not judged, and a consumer beside a held provider is not served the provider's own names raw; mesh-delivery's checker keeps each composition request it sends (state compositions, 3 days) with the read tool checker_composition_requests.

After check c5d32074c945 (issue 1921 notes 69-75), reviewed again: mesh-catalog marks lab moved (lead-47@shanks) and takes in main's mesh-catalog #624 (the fetch downloads each placed sibling's module graph, delivered ahead); mesh-delivery's checker test expects its eleventh state, and it takes in main's #163 (the indexer's take-in waits end, lead-47@shanks) and #164 (a pull request's check composes only after the change's own builds end), both delivered ahead; mesh-host gives the held try's words when it is refused too. Rule bent, said here: mesh-build-go has 52 integration tests, over its budget of 50 (the trunk 49).

Reviewed-by: mesh-code-reviewer@shanks at 67d9ccb1d4

https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

The registry milestone (issue 1966; ADR 0605-0636 as hq's designs name them), taken in onto its trunks after milestone 47/B0 as milestone 47/G's first step (design 47; issue 1921 notes 54-61). One delivery group, branch feat/registry-milestone in mesh-sdk, mesh-contracts, mesh-delivery, mesh-host, mesh-tools, mesh-catalog and mesh-networking: one check, one Approve. No member names another's version (ADR 0573); every mesh package is required at the zero version. This member: mesh-build-go and mesh-knowledge take the trunk: one invokes list (the automatic merge gave it twice), the run step on the delivery SDK's buildstep types and the checker's client, the manifest asked of its owner. The registry's reviewed heads (three rounds, 2026-11-04 to 11-06) were never pushed until now; the check of 068b7d23cf1a judged the heads of 2026-11-04. The trunk step was prepared and reviewed in three rounds (2026-11-07): the merge resolutions keep B0's facts and checking-program rules and the registry's generated types; every check composes with the composer at the node-engine member's head, else its trunk; a module the running indexer refuses but the check's own validator accepts is composed and named in the verdict with its build awaited, never passed silently; the checking program answers only new or changed seat definitions when given the reader's digests, and without them (the running B0 checker) only each seat's own description, 171 KB under the 256 KiB bound. Rules bent, said here: mesh-delivery stays over its test budget (662 unit, 68 integration; was 675 and 68) and mesh-host over unit (1745; was 1746), falling. Out of this step, noted on issue 1921: three asks carry the whole seat list (~430 KB) bounded only by the bus's 1 MiB; the trunk case of the manifest validator is given the heads' seats. After check 616c42434ac2 (issue 1921 notes 63-64), reviewed again: mesh-catalog requires the module of meshdeliverycheckerclient, and a new check checks/requiredmodules refuses an import whose module the go.mod does not require; mesh-delivery's assigner test flushes after its subscriptions (a race), and the checking program's seat answer carries no verb or message schemas (the tool runner's state refuses a schema property named credentials). After check 8a80e8e86858 (issue 1921 notes 66-67), reviewed again: mesh-catalog ignores the new check's program and mesh-build-go's test workspace lays what the composer requires; mesh-host composes from the change's manifest wherever a node's build was made of another (by its manifest digest, or by its contract where the running node-engine names none), naming each such module as not judged, and a consumer beside a held provider is not served the provider's own names raw; mesh-delivery's checker keeps each composition request it sends (state compositions, 3 days) with the read tool checker_composition_requests. After check c5d32074c945 (issue 1921 notes 69-75), reviewed again: mesh-catalog marks lab moved (lead-47@shanks) and takes in main's mesh-catalog #624 (the fetch downloads each placed sibling's module graph, delivered ahead); mesh-delivery's checker test expects its eleventh state, and it takes in main's #163 (the indexer's take-in waits end, lead-47@shanks) and #164 (a pull request's check composes only after the change's own builds end), both delivered ahead; mesh-host gives the held try's words when it is refused too. Rule bent, said here: mesh-build-go has 52 integration tests, over its budget of 50 (the trunk 49). Reviewed-by: mesh-code-reviewer@shanks at 67d9ccb1d455420ee28cd3d9c1afbd769b293e03 https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-tools added 109 commits 2026-11-04 17:50:44 +00:00

CAT-K1: mesh-knowledge asks for its records as a ledger, its states and its reviews stream, and says the new events c973b31820

The manifest keeps postgres-database until the drop and asks, in the typed form, the ledger records (a type for each
kind, the operations of the move), the states drafts, asked, findings and sent, and the stream reviews (kept for good,
refusing when full, granted by the trusted value readers). It consumes claude-code's judged and says the eleven new
events. The mesh-delivery indexer's judge accepts it (checked against main).

Issue 1966.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-C1..C8, C10: mesh-config moves a renamed config value, answers both names in the rename window, retires an unread value on the operator's Approve; the keeper reads its findings 376a8b0fb5

A declaration's renamed-from moves the value with its trust and opens a window per node,
closed on delivered and, for a returning node-engine, on manifest-reads-changed (ADR 0618).
A value no manifest reads is marked retiring only once no node runs a build that reads it,
asked once, removed only on the Approve (a secret through the seat mesh-vault's retire).
New state in the old sections (ADR 0619), the stream config-events, the tools config_renames,
config_rename_again, config_rename_close, config_retiring and the seat verb renames.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-K2: a Store on the ledger, the records' heads without text and each section's text once e0d7644bb6

ledgerStore implements Store on the tool runner's ledger calls: a record is the instance decision/0290 or research/081,
a Save is the operation named by what changed, a section's text is written once in its write-section and named by
digest and sequence in the head, a version is read back from the history, the whole-set findings and the told marks
are keys of the state findings. Every test that runs over each store now also runs over the ledger, played with the
manifest's own types.

The bundle's pipe to the tool runner needs the SDK's raw request (request to A1); until then it answers loudly.

Issue 1966.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-K3: each operation of the records' ledger types has its effect on a head, tested against the manifest's own types 80a7cee6f3

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

mesh-knowledge: a research effort's new version may leave out its sections 285099a940

Closing a long research effort (its status and became links) meant retyping
its whole text through propose. A research effort's new version that gives no
sections now keeps the served version's, and its title, date and deciders when
none are given; the version so composed is judged and reviewed as any other.
A decision, or an effort with no served version, still needs its text.

Issue 1966.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-C9: mesh-config's first whole-module integration test, the rename window and the retirement against a real bus d6c3fc1e40

The built binary, played as the node's tool runtime over a real bus: a renamed value moves and both
names answer, the module restarted inside the window keeps it and says renamed once, a refused move
writes its finding, g14's window closes on delivered, a value no manifest reads is asked about once
when no node reads it and removed only on the Approve.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-V2: a person's rotate of a machine's ledgers secret is refused 4d95d8fc1e

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D1..D5 (holder core): distribution's holding of the two image seats, its caller list, incoming, the store's front and the two windows 9ad22ad1e1

The holder answers publish and publish-part into incoming (parts of at most 1 MiB, resumed from what is held,
hashed, only the missing layers sent, 20 GiB with its 80 % finding and a refusal at the cap, a transfer idle 24
hours dropped), asks builder_versions before it stores, and says published or refused (design 51 *Publishing*,
ADR 0605 §4, ADR 0611). versions, remove (a tagged version; untagged only for the sweep's retention, which a
hold keeps), hold and release (ADR 0617 §2: the sweep alone; the operator's session by hand; never
distribution's own account), tags and mirror. The caller list is a trusted config value checked against the
caller the tool runner proved. The store's port is the program's: reads open, writes refused but while the
direct-write window is open (never over a tag published through the seat) or with the bootstrap's one-off
credential. The old-name and direct-write windows read every node, away ones included, close once, and name a
node holding them seven days.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-K4: a record is written in a draft, one section a call: open, write, remove, order, submit, drop, take-over 17ba95bca6

A draft is a key of the state drafts (author, fields, sections as title, digest and the sequence of their write-section);
write puts the text into the record's history at once, submit makes the draft a version through the rules propose is
judged by and writes no section again. Only the author writes; take-over hands a draft, a proposal or a reserved number
to the caller with the reason in the next version's message. A new version of an effort starts from the served sections
and keeps its title, date and deciders (the rows of the commit folded in, now run over the ledger too). A tool call is
named to the ledger so the runtime stamps the caller it proved.

Issue 1966.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-V1: the seat mesh-vault's retire removes one module's own secret on the warrant of mesh-config's hand act, and says own-retired bb85684b2a

The caller must be on the vault's trusted list retire-callers (the holder of mesh-config alone) and carry
a hand act it recorded for exactly this node, module and name. A waiter on rotated or own-replaced does
not wake on own-retired. Reviewer note of issue 1966: HandedOf returns a placement read error instead of
dropping the modules placed there.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-K5: a section is a part with a plain title, never a file; a reference has one form bd9764ef9a

A new rule, section-form (31 now), refuses from the day the records move a section whose title is a file name, a number
prefix, a slash or over 200 bytes, a title twice in a record, and a text that holds a local path, a file and line, or a
link to a file. Only a section that is new or changed is judged, so a record imported with file-named sections keeps
them until its form-only version; write refuses at once.

Issue 1966.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D1, D2, D3b, D4, D6 (wiring): distribution's program, its tools bundle and its manifest c9b98477b7

The program serves the store's port 5000 in front of the registry container, now on the node's loopback
(127.0.0.1:5001), listed before the program so the port is free when it starts (E14). The tools bundle serves
each seat's verbs as <seat>.<verb> and registry_incoming, registry_holds, registry_old_name and
registry_direct_writes, on distribution's own account, with the caller the tool runner proved. The manifest
claims container-image-registry and oci-artifact-registry with their verbs, keeps mesh-artifact-store and
provides artifact-store until the old name closes (E15), asks its bus state in requires (incoming, transfers,
holds, windows, findings, events) and declares its health outside the program. The bootstrap's one-off
credential is discarded at the first publish answered.

Not yet: the runtime's object store helper for incoming (mesh-sdk), the windows' driver on delivered and
manifest-reads-changed, the read-side health probe calling versions per seat.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-G2 (provision), CAT-G2b: the forge provides go-package-registry and keeps serving the npm registry's public root 4e318c8591

go-package-registry, mesh scope, read with no credential, is offered only by the forge module's new build, so
mesh-build-go's fetch from the Go seat waits for it (design 51, E2); its read side is the forge's Go proxy. The
npm registry's public-root stays served (ADR 0346 §6, the operator 2026-11-04). Both rows of the forge's
manifest test.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-V4, CAT-C10: the vault's retire and the config events and renames verb match the seat definitions of mesh-contracts 553cf50da7

retire takes the definition's warrant (the hand act's id, naming node, module, name and ask) and answers
retired; own-retired, renamed, retiring and retired carry exactly the definition's bodies on its subjects.
The whole-module tests decode what the module published strictly into the generated types. The modules
require mesh-contracts, to be pinned with the group (a local replace in an uncommitted go.work until then).

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-C10, CAT-V1: the modules' readmes say the rename, the retirement and retire e1a182bf09

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-BG3: mesh-build-go fetches the mesh's Go modules through the Go seat first, Go's public proxy second, the forge's git last eeeb2056ae

With go-package-registry bound, the fetch's GOPROXY is the Go seat, then proxy.golang.org, then direct (kept for
this milestone, E11: removed by the first change after mesh-sdk's and mesh-contracts' first numbered versions
are in the Go seat); no GOPRIVATE, and the checksum database off for the forge's paths (design 51 *Go packages*,
ADR 0605 §6). The module requires and binds go-package-registry, so its new build waits for the forge's build
that offers it (E2). Unbound, the fetch is as before.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-RD7: mesh-watcher asks the assigner's seat-holders once and follows its events, no reads 40e1e087e5

CAT-K6: a version is reviewed in parts: the head and each section, a part judged again only when its text or what it names changes c5dc10fa8c

One judgement per part with no tools, at most two at once; a section is read with its frame (the clauses and records it
names and the digest of its Context), so a usual revision judges only what changed. The latest approval of each part is
kept in the head and carried to the next version; a part that could not be judged keeps the others' verdicts and is the
only one asked again; reviewed is said once per version and each verdict as part-reviewed on the stream reviews. A part
in flight is a key of the state asked. A version is accepted when every part approves; findings name their part.

Issue 1966.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-C11: mesh-config reads the indexer's and the assigner's state through their tools and views events 692f18ef3b

Six states (modules, manifests, predefined, config-candidates, placement, wants) are asked of the owners' read
tools at start and kept from their views.<state> events, a gap asks again; a reader started while an owner is
down keeps asking and writes a finding past two minutes. The builder's builds has no owner tool and stays a
read. The returning node-engine's event is read by its node token (ADR 0618 as built). Tools
config_first_views and config_first_view_again.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-K7: every operation's event is sent once, after a restart too 7bf0e991d8

An event of a change carries the record, the version it follows and the event as its message id; the last version sent per
record is a key of the state sent; at start the events of the operations after it are sent again under the same ids, so a
stop between an operation and its event loses none and a second start sends none.

Issue 1966.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-C11, CAT-V3: follow A2's views follower (Changed also says when) 972ecd105d

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-V3: the vault's unanswered first view of the placement is a finding the condition keeper raises 6b929c9057

mesh-vault gains the state findings (old section); past two minutes without the assigner's first view it writes
first-view-unanswered naming the owner and the tool, the answer deletes it. mesh-conditions reads mesh-vault.findings
and holds a row for the kind.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-K8: the search index is in the module's memory (bleve), built from the ledger and kept as it writes a4e3a9cbe0

ask and holds on the ledger answer from a full-text index of the heads and the current versions' text and clauses; it is
built at start (index not built is the answer meanwhile), a new version replaces its record's documents, a reserved,
withdrawn or retracted record has none. A new pure Go dependency, bleve.

Issue 1966.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-V1: retire takes ask beside warrant, as the seat definition now has it 25e5d00aba

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D4b: the program, as distribution's account, discards the bootstrap's credential when the tools close its window b0bfdeb0d1

The tools bundle runs in the tool runner's account and cannot remove a file in
distribution's 0700 state directory, so its first answered publish closes the
window bootstrap-push in the state windows and the program removes the file
(design 51 On a fresh mesh).

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-K9: the move to the ledger: a dry run, the copy as operations, the check, the switch key, the database written beside it 2bf608fc5f

A movableStore serves the database until the switch key says the ledger. The move copies every version of every record as an
import operation (the old author kept apart, marked imported; the ledger's author the module's account), refuses writes for its
length, checks record by record, and sets the switch only on a clean check. From then on the ledger is served and the database is
written beside it, behind or not (database.sequence, database.behind); the check runs at every start before the drop, and the
search index is built then.

Issue 1966.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-K10: the module's tools read each state it keeps on the bus and do by hand what it does itself e5e3d7dce6

knowledge_head, _operations, _section, _draft, _parts, _asked, _fill, _rebuild_index, _review_again, _integrity,
_move and _move_check, beside the health, highest and marks tools; the manifest lists them and a test compares them with the code.

Issue 1966.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-K11: form-only versions: plain titles and links of the one form, proven to change form alone 47b3ae7528

A record imported with file-named sections gets plain titles, and its links to files become references of the one form, in one
version written after the move by the tool knowledge_form_only: marked form, listing the names changed, accepted without a
judgement because a check proves each section's text with its links replaced by their labels is the same and each new link names
the record the old one named. The old names stay as aliases get takes; the parts' reviews carry over; a record waiting for its
review is skipped.

Issue 1966.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-K12: the health on the ledger names the source and the move's last check, and fails for what the ledger owes c73641c117

Unhealthy, naming the cause, for a ledger or the verdicts' stream that does not answer, a head behind its history, an index behind
the ledger and a stream at its cap; the idle drafts, the parts waiting on the judge and the records the form-only check refused
are counted and leave it healthy. A part waiting more than 30 minutes is asked of the operator once, cleared when it is judged.

Issue 1966.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-K13: the knowledge base runs whole on the ledger over a real bus, and a restart finds every record a3a11842bf

One integration test: the module's states are JetStream buckets, its events are published under their message ids to a stream
that drops a repeat; it moves its records, proposes, is reviewed and accepts a record, loses an event between an operation and its
send, restarts, sends the lost event exactly once, builds its index and checks the database within 40 seconds, and is healthy. The
test found that a restart probed the ledger number by number and stopped at a gap of 50, finding 3 of 14 records: the store now
keeps a key held.<kind>.<number> in the state findings for each record it creates and reads them at start (the old probing, with a
gap of 1000, only for a ledger written before the keys).

Issue 1966.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-P1: the overview page's model: three blocks from the ledger and three sources' verbs and events 40b1307839

The model is built once at start from the issue tracker's milestones and list, the delivery groups and the operator channel's asks, and
kept in memory from their events (group, transition, merged; offered, closed; opened, moved, milestone-set, milestone-moved), asking no verb
again; the records come from the ledger and the seven days are counted when the answer is given. The manifest invokes groups and asks,
consumes those events, has no reads and no decided family.

Issue 1968.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-P2: the overview page: a handler that answers GET only, on a port for this machine only, through a required route ab5f767372

The module's bundle (the one process with the bus) serves the page on the port its manifest declares under listens, from: machine,
and requires a route that reaches it. No form, no script, no verb behind it; any method but GET is 405. A source that does not
answer is said on the page, never shown as an empty list.

Issue 1968.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-P3: the page's model is a tool, its re-read a hand act, an unanswered source is said and asked again on a backoff 715bc088c4

knowledge_overview answers the page's model as data (reads); knowledge_overview_reread asks one source's verb again by hand, no other, and
records the hand act (acts). A source marked from its verb is asked again alone from 10 seconds, doubling to 5 minutes, its first answer clearing
the mark; one marked from its subscription asks nothing while silent and the verb once on its next event. The module's health carries the page's
last rebuild and says it is degraded while a source is unanswered, staying healthy.

Issue 1968.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-MS1: the router says when an ask is offered and how it ended 0d86833123

offered once when an ask is offered, closed once for each of the five ends (answered, expired, cancelled by its asker, replaced, ended by the router), with the key, asker and headline and never an option, ticket or proof; on a stream of the events kept seven days, 64 MiB, granted by the trusted value readers, which names the knowledge base alone. Issue 725's runtime change is delivered (resolved 2026-10-26).

Issue 1968.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N1: the bus module's keeper, its own events stream, state and the seat's reading verbs in the manifest 4f24cb0751

The keeper is a process of the tools bundle on its own account (own secret
broker-keeper, ADR 0436 §1): it hears the indexer's manifest-accepted, keeps
what it heard in its old state asked and the persons' requests in requests
(ADR 0595 §10), and says its nine events on its own stream broker-events,
30 days at 256 MiB (design 25 § 12.3, ADR 0620 §2). mesh-broker serves its
four reading verbs; its acting verbs are the keeper's own work. The server
gains a tcp check, which the keeper's tool check stands beside.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N3: the keeper creates each template's bus objects under the module's names and says created or refused ff4d042be5

Each bus-state entry is planned into its objects (design 25 § 12.1): state a
bucket <module>_<local>; events, work queue, log and a bare stream a stream
<module>_<local> on mesh.bus.<module>.<local>.>; objects an object store;
a ledger its history stream, its state and index buckets and its snapshot
stream, one message per subject. The keeper makes what the bus lacks on its
own account; a refusal of its grant waits for the next check, a refusal of the
entry is said (refused) and raised, cleared by a later create. The comparison
of what is held and the orphans are in the same pass (CAT-N4, N5).

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N4: check and drift: a difference that only grows is applied, any other said and raised until cleared 1f68f80760

Rows of the keeper's table test: a larger cap is applied with no event; a
smaller one is drifted with what differs, raised once, and drift-cleared once
the manifest asks what the bus holds again; an entry the bus already holds is
taken as made and its binding said, nothing made twice.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N5: orphans: an empty accepted manifest asks nothing, so what was made is orphaned and kept until a drop 93bfdd9e2f

An entry made for a module whose newest accepted manifest no longer asks for
it is said orphaned and raised, never removed; asked again it is kept, its
binding said again with created and the condition cleared. A module missing
from a first view keeps its key (a set-aside manifest is not in it); an entry
never made leaves nothing, so the empty verdict ends that module's key.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N6: the fill pass raises full at the raise mark and clears it below the clear mark; the keeper's health cc1a942235

The keeper's one periodic read (ADR 0595 §6): every entry that refuses when
full, by its fullest object; full raised at the raise mark (75 by default),
cleared only below the clear mark (70), so an entry at the mark does not flap;
an entry dropping its oldest is never read. HealthOf is the keeper's health
for its tool: unhealthy when its pass is more than three minutes old, its
first view has not answered (naming the indexer's verb), or verdicts wait on
its consumer unanswered past the bound; it names the entries refused, waiting
and full beside.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N2: the keeper's process: the first view from indexer_bus_state, then manifest-accepted on its durable consumer 905c34a537

nats-tools keep is the process keeper, on its own account (own secret
broker-keeper): it asks the indexer's indexer_bus_state once, on a backoff
from 10 s doubling to 5 min while it does not answer (unhealthy meanwhile,
nothing created or orphaned), then binds its durable consumer on EVENTS
(<node>_nats_keeper, the indexer's manifest-accepted) and applies each verdict
only if newer by the indexer's revision; a manifest-accepted carrying no
revision yet is read as 0. A fill pass every minute, waiting entries made
again on the same backoff. It says its events on its own stream and its
conditions through the condition keeper, and serves nothing (ADR 0436 §3).

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N7: the seat mesh-broker's reading verbs list, show, check and orphans, and nats_history 6be2e91da8

Served by the bus module's bundle from the keeper's own state asked and the
bus's monitoring endpoint (design 25 § 12.3): each entry's template, bus
names, settings asked and held and fill against its cap; check names each
object missing or differing; orphans what no accepted manifest asks for.
nats_history reads one entry's events, in order, from the bus module's own
stream BROKER_EVENTS on the tools' credential. The manifest test now holds
every tool and reading verb the manifest names to what the bundle serves.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N8: a person's create, apply, recreate and drop are handed to the keeper as the caller the tool runner proved 3b61a05532

The bus module's tools hold no grant of the seat's acting verbs (design 25
§ 12.3): nats_create, nats_apply, nats_recreate and nats_drop write a request
into the keeper's own state requests, stamped with stdio.CallerOf (a name in
the arguments is never the caller), and read its end back. recreate and drop
raise the operator's ask first, its option binding exactly that act; the
keeper acts only on the router's warrant for it (Warrant.For, Option.Performs),
after a snapshot of the entry's streams (bussnap, moved from the snapshot
command with a streams filter), says recreated or dropped, and records every
act as a hand act through mesh-hand-acts. A drop removes an orphan alone.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N9: the old remove tools refuse an entry of the new form, naming nats_drop 68dd69f4ef

The mesh deletes a module's kept data only through drop and recreate (ADR
0595 §4 as its insight has it; design 25 § 12.3): nats_stream_remove,
nats_bucket_remove and nats_consumer_remove read the keeper's own state asked
and refuse a stream or bucket the keeper made, before the bus is reached,
naming nats_drop and nats_recreate; one that cannot be told apart is not
removed. The old sections' entries are removed as before. Rows of the
existing refusal test.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N6: the keeper's health tool nats_keeper_health, which the process keeper's tool check asks 4b2d5aeaab

It reads the keeper's status from its own state asked, what the keeper holds,
and how many accepted verdicts wait on its consumer <node>_nats_keeper on
EVENTS from the monitoring endpoint, and answers HealthOf's judgement:
unhealthy, naming why, past the three-minute bound; the entries refused,
waiting and past the raise mark named beside.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N2b: nats_first_view_again asks the indexer's first view at once, a hand act 76aab05289

The bus module's tool hands the ask to the keeper through its own state
requests (design 25 § 12.2), as the caller the tool runner proved; the keeper
asks indexer_bus_state at once beside its backoff, applies what it answers,
records the hand act and writes back what it answered and at which revision.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N10: the bus module's one whole-module integration test, against the real bus server fa4ccbc0a9

Design 25 § 12.7's rows, as subtests of one test on one server (CAT-N10): a
verdict, then created on its own stream BROKER_EVENTS, the objects on the bus;
a restart catching up a verdict given while it was down; a replay older than
the first view not applied and a module missing from the first view keeping
its key; a drift said, a growth applied and the drift cleared; a refused
create with no created; a log at a small cap refusing, full raised and cleared;
an orphan dropped only on the operator's warrant, after a snapshot written
whole, recorded as a hand act; nats_first_view_again's hand act.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N10: gofmt the integration test 3e019c5d7e

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N3: the events' body in the field names of the seat mesh-broker's message said cdddeee4ea

module, local-name, storage-type, template, options, configuration, bus-name,
asked, held, fill, snapshot and why as the seat's definition in mesh-contracts
names them, so taking its generated type changes no byte; binding (a ledger's
several names) and revision (the indexer's, of the verdict acted on) beside,
asked of the definition.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-MS1, CAT-MS2: offered and closed carry the operator-channel definition's fields (id, asker, headline, channels; id, asker, how); the page reads them 91cd016f65

The test of the five ends holds every key of a body to the seat definition's holder events, so the router and the contract cannot drift.

Issue 1968.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-CC1: a judgement asked with a ticket says one judged when it ends, with its answer or why it failed 4f433bdaf8

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-K6: a review waits for the agent module's judged event under its ticket; claude_code_judged is no longer invoked 8d4b2cb1c6

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-K6: the manifest test no longer expects claude_code_judged among the invokes d0f384610c

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-CC2: the instructions' seat table reads the seat set mesh-contracts compiles, not mesh-sdk's catalogue e6806d5698

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-CC3: claude-code reads the indexer's modules and manifests and the assigner's placement from their first views and views events, never their state (ADR 0615 §1) 313c1997a4

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N1: the keeper's bus declares keeps, the one key more ADR 0626 §1 allows the bus module's keeper 3daedac1bb

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D3: distribution's mirror copies an outside image by its pinned digest into upstream/<host>/<path> 75986b2960

Until now mirror only answered a copy the store already held, so mesh-build-image had nothing to call for an upstream
base (design 51: an upstream image is copied by mirror with its pinned digest). The copy takes an anonymous pull
token, checks every manifest's and blob's bytes against its digest, copies an index's every platform first, and
fetches only the blobs the store lacks. tags was already served.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N8: the keeper's rows for each refusal ADR 0626 §4 names: a drop naming no ask, a warrant at another level or from another asker, a create no accepted manifest asks for, a snapshot that fails; each said why and deleting nothing ee54bae09f

The keeper already reads the router's record of the ask on its own account and verifies the
warrant itself (bound to this act on this entry, at the destroy level), so only the rows are new.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D4b, D5: the windows' driver moves the old-name and the direct-write windows on start, delivered and manifest-reads-changed 99512eade4

The windows' logic was built but nothing woke it, so both windows stayed open for ever. The program now reads every
node through node_haves and the node-engine seat's manifest-reads at its start, on each delivered and on each
manifest-reads-changed, steps both windows, says their closing once on distribution.events and raises the seven-day
finding. A build module's have counts as publishing through the seats when it runs a build heard publishing (the
tools bundle records the fingerprint of each publish's proved caller), so a paused build agent running the same
build holds nothing open.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D6: distribution's health is a tool the runtime runs outside the program, asking versions through the store's read side 11048b4a2d

An http probe of /v2/ said only that a port answered; design 51 asks the health check to call versions per seat and
probe the read side each minute, naming the program when it is down, from outside its process. registry_health asks
versions of each seat through the program's port, writes the read-side finding while it fails and clears it after.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N7: the keeper says each of mesh-broker's events on the seat's subject, and keeps it on its own stream for history d7214bb576

The lead's ruling of 13:0x (registry milestone): mesh-broker's events are mesh.seat.mesh-broker.event.<event> on the
shared events stream like every seat's, and the bus module's own 30-day stream keeps each for its history tool.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D7: distribution's whole-module integration test against a real bus f8212dc203

The program had no test that runs it whole. This one starts run() on a nats-server of its own, with the registry
container faked on the loopback and node_haves and manifest-reads answered on their call subjects: an away node holds
both windows open, its return closes them, each closing said once on distribution.events, a direct write is then
refused, reads stay open, the health check answers through the program's port, and a waiter on published wakes on
none of the windows' events.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D2: distribution's verbs and events speak the two seats' contracts from mesh-contracts 85fd46b4cc

The verbs answered the holder's own shapes (missing and ended, a part named digest, versions as bare strings, held as
the publish's event) while the seat definitions, now generated into mesh-contracts, name lacks and held, part,
versions of artifacts, tags with digests and a hold's own body. A build module using mesh-sdk's parts helper would
have read no lacking part and sent nothing. The bundle now takes and answers the contract's types and says each
event in its body; an image's manifest is the part whose digest is the publish's, as the contract names no media
type, and a hold on the image seat is of a copied base.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-N7: the seat mesh-broker's reading verbs answer the seat's contract 45688f196e

list, show and orphans answer meshbroker.ListAnswer, Held and OrphansAnswer; check answers CheckAnswer with each
differing entry's settings asked and held by stream and whether the difference only grows; show takes the contract's
local-name. Built against mesh-contracts' seats/meshbroker module in the workspace; the group pin adds its require.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D2: distribution's incoming binds mesh-sdk's seat-neutral receiving side (ADR 0627 §2) 32699e35b8

The holder's own transfer logic goes: parts.Receiver takes the chunks into incoming, checks the hashes, drops an idle
transfer and tells the fill; distribution gives it the builder_versions check, its store and its seats' events, so
the forge's two package seats bind the same receiver instead of a copy.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D5: the publishing builds are written with compare-and-set on the key's revision, so two first publishes at once both count e34fcbeda3

A blind write after a read could lose the build of the publish that wrote first until a bundle restart learned it again.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D3: mirror answers at once and the holder says the copy's end on the seat's events 9f01b9e79a

A large base could outlast a call's bound. The ask is kept in the state mirrors and the copy runs on that state's
watch, so a copy a restart cut is taken on again; published (or refused, with why) on container-image-registry's
events is the end the caller waits on, the place the seat's contract already gives mirror.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

DLV-C3r3 (Go): mesh-build-go fetches a group's tagged version by that version cd66db54bf

ADR 0607 §3, §5: an asked semantic version (v0.1.15-feat-x.2) is downloaded through the Go seat and built whatever
go.mod names, said when go.mod names a newer one; one the seat does not hold fails naming it as a download that may be
retaken, never looked for among the pull request heads. Rows in the asked-version table test.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

DLV-C3r3 (Node): mesh-build-node places a group's tagged version of the SDK from the registry ed7c6c4fb6

ADR 0607 §3, §5: the fetch's package argument also takes {name, version}; the version is placed from the registry in
each TypeScript member whatever its lock names, checked against the lock's integrity when the lock names it, else the
registry's; nothing is packed then, and a version the registry lacks fails naming it. The packed {repository, commit}
form stays for a checker that hands it. Rows in the placement and argument tests.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D3, D2: one mirror copy, said once; a refusal at the cap stays loud until the holder takes a publish again d683a78d9c

A second tools process following mirrors would copy again and say published twice. The ask is claimed by
compare-and-set before it is copied, and its outcome carries the ask's id as the event id, so a redelivery is one
message. The cap's publish-refused finding stands on incoming, not on one file, and clears with the next publish
taken.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D2: distribution's incoming and transfers are its own bus state through mesh-sdk's adapters c6c046283d

The stub that refused every part until the runtime's object store helper existed is gone: a part is now kept in
incoming, a transfer in transfers.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-D2: distribution's caller list and its question to the builder are mesh-sdk's shared parts b4584b0313

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

DLV-C3r3 (Node): name the packed group package the old form, kept while the running checker hands it d31b41bb7e

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-G1, G2: the forge holds both package seats over mesh-sdk's receiving side and caller list ede8994799

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-G1, G2: the forge's Go bundle serves both package seats' verbs into its package registry, for both packages f16efbef4f

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-G2: the forge's npm retention says removed on the seat's events for every version it deletes 0b2260d9fb

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

A3k finding 2: distribution reads its caller list as mesh-config resolves it, not from an environment nothing renders b665156b0f

The node-engine fills ${config:} only in files, unit names, owners and groups, never in a
process's environment, so MESH_CONFIG_CALLERS was never set and design 51's defaults always held.
The tools bundle now reads `callers` through stdio.Config on each call (holder.ConfiguredCallers,
Verbs.WithCallers per call); the program, which has no runtime resolving config, asks its own
tools bundle's new registry_callers on each wake of the windows (manifest invokes). The
environment variable is gone.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-G2: the Go seat keeps what the npm seat keeps; its retention reads every go.mod 4fa457b9bb

go_packages and go_retention read the forge's Go registry as npm_* read npm's: a version a
go.mod on a default branch requires (or replaces to) is kept, with the newest keep; a go.mod
that cannot be read stops a real run before anything is deleted; each deletion says removed
on the Go seat, under the module path and without Go's leading v (design 51 *Retention*).

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-G2: both package provisions carry no credential; the npm pair provisioner is gone 63322ef06f

Reads of the organisation's packages are anonymous and every write goes through the seats'
verbs (design 51 *Publishing*, prepare Q4), so npm-package-registry is form none like
go-package-registry: receives, grants, the consumers' data entry, the grants directory and
the TypeScript pair provisioner with its teams go. Pairs with BN3 (mesh-build-node) and the
lab, which still require the pair.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-G2: the package seats write as an account of their own, the tools' token reads packages 11acf04976

Design 51 keeps each holder's write side as an own secret beside a read-only tool token. The Go
bundle wrote as the forge's administrator; it now reads with the tools' token (read:package added)
and writes with the mesh-packages account's token (write:package alone), which the watcher makes,
mints, keeps and renews.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-G3: the forge's package seats tested whole against a real bus 38ae52e999

The bundle built and played as the tool runtime plays it, its incoming, transfers and findings on
a nats-server of the test, the builder's builder_versions and the config value callers answered:
a numbered Go package goes up in two chunks as the packages account and says published with its
module path; an unnumbered one is refused before the forge and writes the finding. A publish
naming no transfer is now refused: the transfer's name is the caller's own.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-BG1: mesh-build-go publishes a done archive through oci-artifact-registry's verbs d7ffc9cca0

Design 51: only a registry seat's holder writes into its store. The module pushed each bundle
and archive straight into the artifact store; it now publishes through the seat's publish and
publish-part with mesh-sdk's sender, invokes both verbs, and no longer requires artifact-store
or places its address. push.go is gone.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-BG1: the build step answers the reference in the form the artifact seat answers 57a18a4783

ADR 0605 §9: a done key names the seat's reference, oci-artifact-registry/<module>/<artifact>@<digest>,
not the store's older artifact-store:// form; the builder keeps the older form beside it while a counted
node-engine reads only that (ADR 0624). The whole test reads it back as a node-engine routes it.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-BG2: mesh-build-go builds a Go package as its module zip and publishes it under its module path 929b1ce13e

Design 51 Publishing, ADR 0605 §3: a Go package is published from a trunk build through go-package-registry
under the version the builder gave it. The build packs the directory the artifact names as Go's module zip,
answers its digest and h1: sum, and on a version publishes it under the Go module path (the forge's Go seat
takes the name as the module path); a candidate publishes nothing. The ask's seat and version reach the build.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-G2: the packages token is renewed on the forge's refusal, said as a finding, never on a timer 4492bc3013

The Go bundle raises packages-token-refused in the module's findings state when the forge
answers 401 to a write; the watcher mints afresh on that key's event and clears it (lead's
ruling on A3n, map-onto-how-the-mesh-is-built: nothing waits on a timer).

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-BG2: the step answers no h1: sum, since nothing in the contract carries it on 6146c8a76d

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-REG: distribution's references name the seat as their registry part cb3db6c902

The reach came from MESH_REGISTRY_REACH, which nothing sets, so every published event
answered /<name>@<digest>. The store answers the path; the holder puts the seat's name
before it, as catalogue.InRegistrySeat reads it back (ADR 0605 §9; lead's ruling on A3n).

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-G2: gitea reads no MESH_* word nothing sets 5450707aaf

The forge's container is the manifest's name, and the hand-run token is GITEA_TOKEN
alone (checks/environment).

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

checks/environment: refuse a module reading a MESH_* word nothing sets aaca1c6ca9

A word its manifest does not set and the mesh gives no process is a contract nobody
keeps (MESH_REGISTRY_REACH, MESH_CONFIG_CALLERS). The 63 reads outside this milestone's
modules are listed in unset, which only shrinks (lead's ruling on A3n).

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-BG5: mesh-build-go regenerates a contracts module's Go files and names each out of step c5bec10bb5

The unit level fails on a file that differs, is missing or was not generated; a build
ask answers the same files in its out-of-step, a failed build's too, so a trunk build's
end raises generated-types-out-of-step (lead's ruling on A2k finding 1). The compare is
buildstep's, each build module its own root, with the seat generator as a requirement.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-BN2: mesh-build-node regenerates a contracts module's TypeScript and names each file out of step b5495d0cc8

The unit level fails on one, a module with no tests of its own too; a build ask answers
them in its out-of-step through buildstep's compare (lead's ruling on A2k finding 1).

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Join: merge A4 (wip/registry-a4) into the registry milestone 6061c775d5

CAT-BG4/BN1: the seat generator moves into buildstep, one program for Go and TypeScript 7b08f25d4a

Its version is the build module's own, so a new generator is a new build of
mesh-build-go and mesh-build-node and a new fingerprint of what they compare;
no repository of its own. Moved from seatgen c1fdd7c unchanged but for the
import path.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-BG2: a bundle or archive the builder numbers later is published late, sending nothing d9d1119a72

mesh-build-go consumes the builder's version.numbered and publishes the held
digest under its version, naming its one part by digest only, since the packed
bytes left the node; distribution tags it at the size its store holds, so the
version's manifest names the blob truly.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-BG6: buildstep is declared as mesh-build-go's Go package, series 0.1 39a8d8d7ee

So the builder numbers it and a user outside this repository accepts a series
instead of naming a commit.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-BN3: mesh-build-node publishes through the registry seats and holds no registry credential af280050d6

Bundles go through oci-artifact-registry and answer the seat's reference;
a trunk package is packed under the builder's version and published through
npm-package-registry, so the holder can check it with the builder. Reads are
anonymous, so the write pair, the sealed secret and the store's address go.
The one-part sender moves into buildstep for both build modules, so the move
to the registry clients is made in one place.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-LAB1: the lab clones through the forge account, and holds no registry secret e40ace34a6

npm-package-registry is form none, so the registry pair the lab's clones of
mesh-tools-go used is no longer minted. The lab requires forge-account as the
build modules do; its clones go over ssh through a git and ssh configuration the
mesh writes from that binding. The secret, contributes and MESH_LAB_FORGE_SECRET go.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-BI1: mesh-build-image publishes and copies through container-image-registry d47d231ec3

A trunk image is written as an OCI layout and published by the seat's
publish and publish-part, only the parts the holder lacks; an upstream copy is
the seat's mirror, ended on the holder's published or refused. The module
holds no registry address: artifact-store, files/registry and the copier go,
and the step answers the seat's reference.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-PB: buildstep.PushBlob is removed, its last users write through the seats 6da45bc090

No build module writes into a store any more: mesh-build-go's store test
plays the holder with its own two-request upload, as mesh-build-node's does.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-BI1: a copy keeps its wait for the mirror's outcome in the state mirror-waits 1c4739698a

The image seat's published and refused reach one instance of the module, not
necessarily the one whose copy waits, so an outcome held only in memory was
dropped. The instance that hears it now writes it into the copy's key, and the
copy reads it there, woken by its watch of the state; build_image_mirror_waits
reads the state.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-NR1: every build narrates its asks of a seat in one place 37a1f2be53

asked, waiting on the seat every five minutes, stored or refused, as the ask's
phase on its step key, so a publish or a copy that stands still is readable on
the bus. PublishOne and PublishFiles narrate for every build module; the image
copy and the late publish use the same narration.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-AS1: the artifact sweep holds and removes through the registry seats 4915e67520

The store's write side is its holder's now, so the sweep asks the seats'
hold, release and remove instead of writing to the store's door. Both
reference forms meet as one artifact; a refusal stops the sweep and writes
the finding sweep-refused, which a clean sweep clears.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-AS1: the store's door writes nothing; its write code goes 37a16a4e98

The sweep holds and removes through the registry seats, so the store's
LetGo, holds and holder manifests, and the unused GetBlob and
HoldsManifest, have no caller. One table test reads the door's two
remaining answers in both reference forms.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-AS1: distribution removes an index's platforms with it, but those a kept index names b88c6dac5b

The sweep asks remove now instead of deleting over the store's door, so
what its door did for an index is the holder's: a removal for retention
removes the platform manifests no tagged or held index of the repository
names, and refuses a manifest such an index names.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Commit the build modules' publish code a bare ignore pattern hid; drop committed go.work.sum eb50bf1f48

The patterns mesh-build-node and mesh-build-image matched the command
directory, so publish.go and its test were never committed and the
branch did not compile. Anchored to the binaries now.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Build modules publish through the registry clients over the parts sender 348f9631fe

mesh-sdk's publish package left mesh-sdk (ADR 0627): buildstep binds each
registry seat to its client in mesh-contracts, over mesh-sdk's
seat-neutral sender of parts. The late publish uses the same path.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

CAT-BI1: mesh-build-image publishes a numbered image's held digest late (ADR 0617 §1) 31f6c635f5

An image built unnumbered is published by its digest; the builder's
version.numbered now publishes it under its version, sending nothing.
The late publish moved into buildstep, so mesh-build-go and
mesh-build-image share one.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

seatgen: a string's enum is one Go constant per value 9c8d64e28c

A caller names a value the seat definition gives, such as the removal
reason retention, and never writes it by hand.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

A release of nothing held answers released false; the sweep names retention from the seat 4438045b8c

distribution answers released false for what it does not hold, as the
registry seats now define, and takes the removal reason from the
generated constant; the sweep reads the answer instead of matching
the error's text.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

buildstep requires the registry clients and seats it imports 9d09aeddb1

Placeholder versions, as for the other contracts modules; the group's pin
writes the real ones.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

mesh-tools commented

2026-11-04 17:51:07 +00:00

Delivery novox/mesh-catalog@67d9ccb1d455 — ready since 2026-11-07T13:39:24Z

Group feat/registry-milestone: novox/mesh-catalog@67d9ccb1d4, novox/mesh-contracts@b77560786a, novox/mesh-delivery@4735fc0b5f, novox/mesh-host@cb7a444a56, novox/mesh-networking@9ee0ffe3c0, novox/mesh-sdk@d932546703, novox/mesh-tools@6c05e42626

Transitions

2026-11-07 13:39 (new) → proposed (announced): novox/mesh-catalog#612's head announced
2026-11-07 13:39 proposed → ready (accepted): a verdict names this commit and the trunk's head as its trunk commit, the gate passed or warned, and the repository's own check did not fail; or the check gates no merge (check-gates-merge false)

The commit's note under refs/notes/mesh-plan keeps every transition of this change up to its merge: git log --notes=mesh-plan.

**Delivery** `novox/mesh-catalog@67d9ccb1d455` — **ready** since 2026-11-07T13:39:24Z **Group** `feat/registry-milestone`: novox/mesh-catalog@67d9ccb1d455, novox/mesh-contracts@b77560786a6a, novox/mesh-delivery@4735fc0b5fdb, novox/mesh-host@cb7a444a5634, novox/mesh-networking@9ee0ffe3c0b6, novox/mesh-sdk@d93254670307, novox/mesh-tools@6c05e4262641 **Transitions** - 2026-11-07 13:39 (new) → proposed (announced): novox/mesh-catalog#612's head announced - 2026-11-07 13:39 proposed → ready (accepted): a verdict names this commit and the trunk's head as its trunk commit, the gate passed or warned, and the repository's own check did not fail; or the check gates no merge (check-gates-merge false) The commit's note under `refs/notes/mesh-plan` keeps every transition of this change up to its merge: `git log --notes=mesh-plan`.

mesh-tools commented

2026-11-05 23:47:35 +00:00

mesh-delivery: the operator's Approve of this pull request's merge into main is ended: a head it covers moved on: the heads it covers are no longer the group's heads. Nothing is merged on it. The operator is asked again once the checks pass at the new head.

mesh-tools commented

2026-11-07 00:32:41 +00:00

mesh-delivery: not merged. The merge of this pull request into main at 125929dbc4 was declined: the operator, via telegram (user id verified), chose Decline. A new head asks the operator again.

mesh-delivery: **not merged.** The merge of this pull request into main at 125929dbc459c2e278e44630bf14d7edff16a77e was declined: the operator, via telegram (user id verified), chose Decline. A new head asks the operator again.

jschoubben added 175 commits 2026-11-07 02:54:16 +00:00

CAT-RD8: minio asks the assigner's seat-holders once and follows its events, no reads, no cache on a timer aff9d6ae6a

CAT-RD9: nfs-server asks the assigner's seat-holders once and follows its events, no reads 9e726d4c41

CAT-RD6: mesh-self-check asks the owners' placement, seat-holders, modules and manifests once and follows their events, no reads 5bc05201ef

CAT-RD1: artifact-store-tools asks the assigner's wants and the indexer's manifests once and follows their events, no reads 2339ac5af7

CAT-RD4: mesh-conditions asks the assigner's wants, placement and findings once and follows their events, no reads (on CAT-C10's base) f62c2ff380

CAT-RD5: mesh-provisioner asks the assigner's placement and wants and the indexer's modules and manifests once and follows their events, no reads 550bb4fdd1

CAT-RD3: mesh-build-go asks the indexer's read tool for the manifest a compose ask names, no reads of the indexer d618cbb9ca

CAT-RD2: gitea asks the indexer's modules once and follows its events (TypeScript follower), no reads; mesh-build-go's manifest text kept as it was b00a826681

CAT-RD10: the deprecated mesh-catalog module reads none of the indexer's state; it registers builds without manifests, said once (ruling U4) 7843edb1eb

CAT-RD1..RD10: the readers use mesh-sdk's views.Reader, say an unanswered first view in their own findings (mesh-conditions observes them) and record a re-read through mesh-hand-acts 4d597eb9dc

CAT-RD2: gitea says an unanswered first view in its findings and records the re-read through mesh-hand-acts bcac2d6d1c

Join: merge A8 (wip/registry-a8) ea4e22577c

Conflict in mesh-conditions keeper.go Observers and module.json reads: origin/main
added mesh-vault's findings, A8 added the first-view readers' findings; both kept.

Join: merge A5 (wip/registry-a5) c295176048

Join: merge A6 (wip/registry-a6) aad5c6ceea

Conflict in mesh-knowledge main.go, the propose tool's description: #609 (79c2fc7f)
names ideas and the research-active-at-most limit, A6 names a research effort's
partial new version; both sentences kept, A6's rule count (31).

Join: merge A2 (wip/registry-a2) a46e52000b

Join item 3: the condition keeper reads every finding the registry milestone added 87a30055a6

The builder's four kinds, the indexer's four, the package seats' token refused and the
owners' view too large get their homes; the keeper reads the findings of every reader
of the indexer's and the assigner's state (teller, enrolment, builder, checker, view),
and asks the indexer's findings of its read tool and follows its views events, as the
assigner's (ADR 0615 §1). Rows in TestEveryControllerKindHasItsHome and the manifest
test; each failed before the change.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Join item 1: the catalogue's modules take the router's ask from operator-channel's contract and client 9d83987815

mesh-sdk holds neither asks nor asker, nor the vault's answer types, and mesh-delivery's
sdk/views is mesh-sdk's go/views (a1-progress.md, the rewrite map for the join): thirteen
modules switch to the operatorchannel contract and client, mesh-sdk's channel package and
meshvault, with go.mod requiring each at the group pin (nats: meshbroker too). The gate
holds the checks mesh-sdk dropped (an ask's id is its key's asking, at least one option),
with a row that fails without them; claude-code refuses a router record whose times do
not parse, as its JSON decode did, with a row.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Join item 2: mesh-knowledge's ledger pipe, call ids and page heartbeat come from the SDK 2770a55b61

The ledger pipe asks the runtime through stdio.BusState (it answered ErrNoPipe before), an
operation names the call the runtime proved (stdio.CallIDOf), and the page takes the
runtime's word on its consumer (stdio.OnHeartbeat): a beat is every subscription heard, a
missed heartbeat marks each subscribed source at once with the runtime's why. Rows in
TestASourceThatDoesNotAnswerIsSaidAndAskedAgainOnABackoff and
TestTheLedgerStoreRefusesWhatTheLedgerWouldHaveToTakeBack; each failed before the change.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Join item 1: claude-code requires the mesh-contracts modules it imports, for the group pin 2c807a3800

Its seat set (mesh-contracts/go, A6's CC2) and the operator-channel contract and client
were imported with no require; the root workspace hid it.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Join item 6: gitea announces its findings readable, which the condition keeper reads 6ec562e964

A8 gave gitea a findings state and the keeper its read; gitea's announcement did not name
it, so the read was refused on the mesh (checks/readable) and its own kept test failed.
Rows in test/kept.test.ts; they failed before the change.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Join item 6: no compiled program in the tree, and the health count written down 880de92bc8

A8's CAT-RD4 and CAT-RD8 committed the mesh-conditions and minio binaries
(checks/nobinaries): removed and ignored. The long-running resources without health fell
to 57 in the joined tree (checks/health ratchet).

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Join: mesh-config's whole test reads the subject the runtime publishes on from the manifest 3f39b60d9a

The seat definition names mesh-config's renamed, retiring and retired on its own stream
(mesh.stream, A1c), and the runtime publishes an event the module's stream carries
there; the test assumed EVENTS (mesh.mod) for every event and failed at the joined heads.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Join: mesh-knowledge's progress bus test counts the page's first view 725eea8a68

At the joined SDK the bundle is served in the test, so ServePage's first view asks the
tracker's list and milestones and mesh-delivery's groups once beside progress (checked:
with the page's start skipped the counts are the old 3, 1, 3).

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Join: merge origin/main (#610 , the knowledge base's close verb) e71b006d85

Conflicts in mesh-knowledge: main.go Verbs and manifest_test.go, A6's draft verbs and
#610's close, both kept in the manifest's order; review.go reviewOne, A6 judges a version
in parts and #610 judges a closing on its closing alone: a closing version is one part
(ClosingPart) asked ClosingQuestion. Joined: close.go's asks import by the rewrite map,
and its rules probe names its overview section "Overview", since A6's section-form rule
refuses a file-named section (close's test failed on it).

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

J7: the gate reads the ask it keeps through operator-channel's client (occ.Carried) 05dd17ebaa

askOf keeps its key step and holds the carried ask to its id and an option with
occ.Carried, the one reader the assigner, teller and view use; its own decode, id and
option checks are gone. Rows in TestTheGateRefusesAnAskWithAnotherIdOrNoOption name the
client's refusals and a value carrying no ask; they failed with the gate's own checks.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Join: merge A3 (wip/registry-a3) d3d4dbc8ed

Conflicts, both sides kept: artifact-store-tools' invokes (the assigner's
and indexer's read tools, followed through views, beside the registry
seats' hold, release and remove), its main (the views reader and the
seat asker), its manifest row; gitea's invokes (the indexer's read tool,
the hand act and the builder's versions) and its consumes of the
indexer's views. gitea's account test takes the packages account's
names; repoCheckFacts, removed on main, is not imported. No catalogue
module reads the indexer's state.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Join: merge origin/main (#611 , a closing reviewed on its closing alone) ad671183bf

Conflict in mesh-knowledge's reviewOne: the join judges a record in
parts, #611 asks a closing only of the records its judged links name,
with the status it moves from. Both kept: a closing is one part,
ClosingPart, asked #611's ClosingQuestion over the named records.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Join: mesh-knowledge's page address is no word of the environment b4e7ffd89c

The overview page read MESH_KNOWLEDGE_PAGE, which nothing in the mesh
sets; only its bus test did, to serve no page. checks/environment,
which came with A3's merge, refuses such a read. The page listens on
the port its manifest declares; the test's bundle sets the package's
address empty instead.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Join: mesh-build-go's manifest gives invokes once; gitea's row names all three fd908a444d

A3's merge auto-merged a second invokes into mesh-build-go's manifest
(the registry seats' publish verbs beside the indexer's read tool); a
decoder kept the last and dropped the read tool, which
TestTheModuleReadsTheIndexersManifests caught. One list holds both.
gitea's package-registry row named only the builder's versions; the
manifest also asks the indexer's modules and records a re-read.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

checks/manifests refuses a manifest giving one field twice d3fa102e4d

A decoder keeps the last of two equal keys without a word, so a merge
of two branches that each add a field to a manifest loses one side
silently; the indexer's judge reads the same bytes. Rows: a top-level
field and a field inside an artifact, each named by its path.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

claude-code: drop a committed go.work.sum, a local workspace leftover 65c2d76506

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

F-catalog-B: distribution removes a tagged version's tag alone when another tag names its digest 1f79a4d52a

Deleting the manifest by its digest removed every tag on it, and reproducible builds share a digest
between a branch's tagged version and the trunk's numbered one, so a branch's end deleted the trunk's
image (security review F1). The fake catalog now deletes by digest as the store does; a store that
cannot delete a tag alone removes nothing and the removal is refused.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

F-catalog-B: distribution's front takes no bootstrap credential 7cd9b611d2

The lead's ENG-5 ruling: on a fresh mesh the bootstrap pushes the holders' own images to the store on its
loopback with no credential, so the front's bootstrap token, its file, the bootstrap-push window and the
program's removal of the file go. Fewer secrets; nothing to refuse once a window closes.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

mesh-config: the whole test waits on the module's views subscriptions, never a second's sleep 5ef6287ce4

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

F-catalog-B: distribution's windows step by mesh-sdk's delivery.StepWindow and say why a node holds one 39cd545814

A read naming no node (an empty node_haves) closed both windows for good (security review F3, second of
a kind): the window embeds delivery.WindowHeld and steps by the shared rule, named being every node the
read named, and keeps its own guard that a window once closed stays closed. A node-engine whose
manifest-reads did not answer (a timeout included) counts as stating nothing as before, and the window's
record now says so per node (Why), so registry_old_name tells a timeout from an old node-engine. A read
that fails otherwise is an error the wake judges nothing on, no longer taken as not served.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

F-catalog-B: distribution counts incoming against its binding's cap and drops an idle transfer by its expiring key b4ab61c0aa

The holder counted against 20 GiB while the bus caps incoming at 8 GiB, so the bus refused writes before
the holder said refused naming the cap, and the 80 % finding came too late (running review F2). The cap is
read from incoming's binding (the object store's fill); IncomingCap, 8 GiB, is said only where none is read,
and the manifest's row holds it equal to max-mib. An abandoned transfer was dropped only on the next publish,
and the transfers state's own lifetime dropped its key while its parts stayed: transfers keeps for good, and
each open transfer has a key in the new state idle with its own time to live, whose end drops it (re-armed
for the rest when it took a part since). registry_incoming reports the bound cap.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

mesh-conditions: observe node-tools' and claude-code's first views and the builder's publishing kinds 4c4abb1b66

Their views findings were written and raised by nobody (ADR 0615 §6). The test finds every catalogue module
writing a views finding in the tree, so a new writer cannot go unobserved; homes for publish-refused,
publish-unanswered, seat-outcome-unread and family-not-heard.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

F-catalog-B: registry_mirrors reads distribution's mirrors, asks one again and drops one by hand 6cd4aedb7c

The state mirrors had no read tool and no ask-again (seams review note 1): a copy a follower holds and
does not end could only wait for a restart. registry_mirrors lists each kept copy and its claim, asks one
again with its claim let go, and drops one, saying refused under the ask's id so its waiter ends.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

F-catalog-B: npm_incoming and go_incoming read and drop the package seats' transfers; an idle one drops by its key 7a60771165

The package seats kept incoming, transfers and findings with no read or drop tool (tools review F4).
npm_incoming and go_incoming list a seat's open transfers, incoming's fill and the cap its binding carries,
and drop one by hand. The same fault as distribution's (running review F2, its second instance): the
transfers state's own lifetime dropped a transfer's key and left its chunks, and the drop otherwise waited
for the next publish. transfers keeps for good; each open transfer has a key in the new state idle with its
own time to live, whose end drops it. The holder counts against its binding's cap.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

F-catalog-B: the gitea module removes the npm pair's teams packages and readers and their accounts ce989b001a

ADR 0605 §5 revokes the npm write token, and the 14:04 deferral is withdrawn (security review F5): the
accounts the removed pair provisioner made (mesh_<node>_<consumer>) still logged in with write on the
organisation's packages. cmd/agent-teams, run at each start of the module and change of its teams value,
removes both teams and each account of that form whose only teams they are, never purging: one the forge
refuses to delete because it owns something keeps it and has its login prohibited.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

F-catalog-B: no ignore pattern hides a module's source; nfs-server's and restic's patterns anchored cd3e444a76

A bare pattern meant for a binary (mesh-build-node) matched the command directory of the same name, so git
ignored every new file there and publish.go was never committed (catalogue review F1). The kind: every
bare pattern of a module's .gitignore is anchored to the binary (nfs-server's hid all of cmd/nfs-server),
and checks/gitignored refuses an ignore pattern, the module's or the catalogue's, that hides a directory
under a module's cmd/ or internal/, a Go file added there, or a source file there, in a git checkout and a
plain folder alike.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

F-catalog-B: the catalogue's Go is as gofmt writes it, and checks/gofmt holds it so b5644829d3

buildstep's generated.go and generated_test.go and claude-code's config_test.go reached the join
unformatted: the second of a kind, which no check caught since each module's tests hold only themselves.
All three are formatted, and checks/gofmt refuses a Go file of the catalogue gofmt would change.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

nats: the keeper wakes on decided, spends a warrant once, and says again what the bus did not take 1af1a7b002

Review round 1 of the registry milestone (design 25 § 12.2, § 12.3, § 12.7):
- the fill pass runs from the start, while the first view is asked;
- recreate and drop wake on the operator channel's decided on the keeper's durable consumer, never a minute's tick;
- a warrant is spent by compare-and-set (acted.<ask id> in requests) before anything is deleted; a second request on it
  is refused; a request whose ask ended unanswered is ended, said; the ask is raised before the request is handed;
- created is marked only once said; an event or a condition the bus did not take is kept, named by the health and
  said again at the next fill pass with its one id, on the subjects left (the double publish);
- the health judges the first view past its bound and never skips an unread consumer backlog;
- an unrecordable hand act is said on the request; nats_requests reads what waits, what ended and each warrant spent,
  and drops a waiting request by hand; a tool waits for its request's end on one watch of requests;
- the whole-module test reads the warrant from the router's record under a process connection.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

F-parts: distribution and gitea bind mesh-sdk's idle key and binding cap instead of their own copies 34269df920

Both holders carried the same idle bound and the same read of incoming's cap; the second instance of the
kind moved into mesh-sdk's Receiver (Arm, Expired, Bound, with BusArm, BusCap and FollowIdle). Each
holder now hands its idle state and its objects to those, and keeps its rows.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

F-parts: on registry 2.8.3 a shared-digest tagged version's remove keeps the tag and says removed ddcd790304

The store's pinned image is registry 2.8.3, whose DeleteManifest cannot delete a tag alone: by a tag it
deletes the empty digest and answers 400 DIGEST_INVALID. The fakes accepted a DELETE by tag, a fixture
agreeing with the mistake, so every merge of a reproducible build would have been a refused remove and a
retry finding live. RemoveTag now answers ErrNoTagAlone for that answer; remove then keeps the tag, whose
bytes stay under the other tag anyway, and says removed with that reason, so the builder lets it go. Any
other store failure is still refused; a digest shared with nothing is still deleted by digest. The fakes
answer as 2.8.3 does.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Merge branch 'wip/registry-fcat-b' into feat/registry-milestone 4b93dbb04e

Pin mesh-sdk c54648e74346 and mesh-contracts ec9529022c44, the group's heads, in every Go module (issue 1966) 80cdcbfa2f

Every module that imports mesh-sdk requires its group head; each requires the mesh-contracts seats and clients it
imports, and mesh-contracts/go through the SDK's seat list. The placeholder versions are gone.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

G-catalog-B 5: distribution and gitea look at every open transfer at the idle watch's start dad2708805

A transfer whose idle key ended while the bundle was down was never dropped. Both hand mesh-sdk's
FollowIdle their open transfers, which it gives Expired once the watch is taken.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

nats: a warrant's spend record in the keeper's own state spent; the fill pass while the consumer binds; partial says kept f86a1f2a36

Review round 2 of the registry milestone:
- security F2: the spend record leaves requests, which the tools write as the tool runner's account, for state spent,
  named in the keeper's bus.state and declared read-only (the lead's 18:1x ruling: the tools read it, never write);
  nats_requests lists the warrants spent from it;
- running F1: the fill pass also runs while the manifest-accepted consumer is bound again;
- health note 3: an event said on the seat's subject and refused by its history is kept with its one id and said again
  on the history alone, never again with a new id;
- seams note 3, health note 4: words unsaid with a status that cannot be written raise the keeper's own condition
  status-unwritten, cleared when it is written; the health names what a restart would lose.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

G-catalog-B 7: a transfer dropped by hand is recorded through mesh-hand-acts first ac29f20517

registry_incoming, npm_incoming and go_incoming dropped a transfer by hand with only a log line. Each
drop now records its hand act through the seat mesh-hand-acts before anything goes, as nats_requests
does; an act not recorded leaves the transfer open. distribution's manifest invokes the seat.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

G-catalog-B 8: the retire removes only the teams the npm pair provisioner made 480556a329

A person who later makes a team named packages or readers would have lost it at the next start. A team
counts as the provisioner's only in its organisation and when every member is one of its accounts (it
put its own accounts there alone, and they go with the first run), so a team made later stays.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

nats: the manifest test reads spent read-only through the SDK's own decoding f3983ed473

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

G-catalog-B 6: on registry 2.8.3 a shared-digest tagged version's remove answers and says kept 4a3d7b60a1

Saying removed for a version the store still holds told the builder something false. remove now answers
kept with the reason shared-digest and the other tags naming its digest, removed false, and says the
seat's event kept with them (mesh-contracts' shared registry part); the builder raises no retry.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

mesh-conditions: a home for view-unsaid, every views owner's own c275919f11

mesh-sdk's views.UnsaidFinding (id <owner>.<state>) is written by the indexer and the assigner in their own findings;
the closed set finds a catalogue writer of it too. nats.status-unwritten is said to the keeper directly, not a finding:
no row.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Merge branch 'wip/registry-gcat-b' into feat/registry-milestone c50a7d059a

Pin mesh-sdk e5d89c5eedb5 and mesh-contracts 347a3053580a, the group's heads after review round 2, in every Go module (issue 1966) 5ab349df6c

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

mesh-vault: secret_own's description says it lists the own secrets of a machine, so a search for those words finds it (issue 1966) 1f9b130225

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Pin mesh-sdk 05b3e2e1b94f and mesh-contracts ae5266e77639, the group's final heads, in every Go module (issue 1966) d8af436536

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Ask git ls-tree for nested named paths in rounds: given go and go/seats/x together it lists go's children, so mesh-contracts' preview failed (issue 1966) 680da05c4c

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Ignore every program a local go build leaves, and refuse a main package whose program is not ignored (issue 1577) 4d3c901b99

Twice in the registry milestone a module's program, built in its folder, was committed with git add -A
and stayed in the branch's history. The ignore list was kept by hand and covered 56 of 229 programs.
checks/gitignored now lists every main package's program (in its directory and at its Go module's
root) and fails on one no ignore file names; the root .gitignore names each, anchored to its path.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Keep Go's indented continuation in a step's cause: the composer's reason ended at "requires" and hid the module that could not be found (issue 1966) d0a69afbb1

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

mesh-vault: the player holds a family's events until the vault subscribes, and the rotation test waits for the vault's placement b4e0518282

The rotation test failed twice under parallel runs ("never: the rotation begins"), from two races in
its player. The player refused mesh/subscribe, so the placement events it said never reached the
vault: when the vault's first view was answered before the test wrote the placement, the vault never
knew the bus node. And the test wrote the bus node's have right after the placement, so the vault
could hear the have before it applied the placement, adopt nothing, and never learn the loaded list's
revision. The player now answers mesh/subscribe and holds a family's events until it is subscribed,
as the module's consumer does; the test writes the placement only after the vault's first view and
the have only once the vault's picture holds every placement key (vault_first_views), so both orders
are fixed rather than waited out.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Make five manifests the milestone changed pass the indexer's judge: an event in one place, one form per state, a digest of a text field, trusted and runtime config values (issue 1966) cea09d8a36

messenger's refusal took its declaration of operator-channel out of the judged catalogue, so claude-code's use of it
was refused too. distribution's two remaining refusals (its program builds beside the artifact store it provides,
and its only health check is its own tool) are a design question left to the lead.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

mesh-vault: adopt from the bus node's kept have when the placement names it af487d38ed

A have of the bus module heard while the placement did not yet name its node as holding the bus
adopted nothing, and nothing adopted it again until the have changed, so a vault could stay without
the bus node's adoption and the loaded list's revision for good. The vault now keeps each node's
latest have and, on the placement's next change, adopts from the bus node's kept have if it was not
adopted: the placement's event, no timer. The rotation test now writes the have before the
placement (its wait for the vault's picture is gone), so it takes this path every run.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

distribution: the store's /v2/ checked over http through the front, beside the program's tool check (issue 1966) 3dfc3533be

A tool check alone is refused by the indexer's judge (ADR 0227 rule 8, ADR 0240 rule 2); the operator chose a check of
another kind on 2026-11-04.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

artifact-store-tools: require no artifact-store; the sweep reaches the store through the registry seats (issue 1966) 10bbbadcf2

The operator's decision of 2026-11-04: the store's clean-up asks the registry seats' verbs and is no consumer of the
provision.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

health-undeclared: 56, the artifact store's container now says how it is ready (issue 1966) 64cb7ecc53

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

distribution: provide and serve no artifact-store, and answer the seats' list from the store's own files (ADR 0632 §1, §3, §4; issue 1966) 541c2e02d9

The store is found by the seats' holder and its listen registry; list says,
per repository, each manifest's digest, whether it is an index and what it
names, its tags, its size and whether it is held, so the sweep reads the
store through the seat. E15 now asks that nothing is provided or served.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

artifact-store-tools: read the store through the registry seats' list, never its files or its read side (ADR 0632 §1, §4; issue 1966) 1bf98663c6

The docker exec listing and the HTTP door go; held and the sweep's index
skip ask the seat's list too. The store's address file is the image seat's
reach, used for the catalogue's definitions only.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

docker: trust the store at the image seat's reach (ADR 0632 §1; issue 1966) bbb9e37df4

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

mesh-conditions: the indexer's manifest-held has its home (ADR 0632 §2; issue 1966) 236ce2fe12

A module's new entry held from the whole mesh past its bound is a finding of
the indexer's; without a home the keeper would not raise it.

Claude-Session: https://claude.ai/code/session_01JG73KG7AfSdh6FP8VrBqCn

seatgen: a seat's states give a constant, key builders and parsers and strict value types, Go and TypeScript (ADR 0634 §6; issue 1966) 0beef7ff9c

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

seatgen: no seat list and no root package, definitions in any folder, seat-list-grain files generate nothing, packages with a dependency-free lock (ADR 0635 §1, ADR 0634 §8, F7; issue 1966) 1abae72457

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

mesh-build-go: the generated-files compare finds a contracts module in any folder; a folder module's Go packages each publish under their own module path (ADR 0634 §8; issue 1966) 2b2b3e4b8c

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

mesh-build-go: a real-git test that a root module's source leaves out its nested module folders and a folder module is narrowed like a root one (D-I4, ADR 0634 §8; issue 1966) a2272916a0

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Merge mesh-catalog main (559939aa) into the registry branch: gate reads forgemsg, messenger narrows asks (K-0; issue 1966) 518539f3d1

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

Mesh-catalog modules import operator-channel from its domain's contracts module and the registry seats from mesh-delivery's (K-4..K-20 first part; ADR 0634 §7, ADR 0636; issue 1966) a86e2ff1fc

The group's module paths are placeholders the re-pin rewrites.

Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig

messenger: the asks it says are a bus-state entry of the events template, not a stream of its own (K-24, F6, ADR 0595, ADR 0619; issue 1966) 7ae2ed88c0

claude-code: the seat list is read from the seat mesh-catalog and followed on its events, the hand acts go through the hand-acts client, the seat is mesh-module-assigner (K-23, K-21 part, K-4..K-20 part; ADR 0635 §1, ADR 0634 §9; issue 1966) 2f7ff1c911

The seats are called mesh-module-assigner and mesh-node-enroller in mesh-cli, claude-code, the checks, mesh-config, mesh-vault, conditions, self-check and the provisioner client (ADR 0634 §9; K-21; issue 1966) 9fafbcfd58

Eleven modules record hand acts through the hand-acts client, the finding writers are the conditions client's, nats takes the broker seat's name from its package (K-21 part, K-4..K-20 part; ADR 0634 §4, ADR 0635 §4; issue 1966) 87388b9046

Merge b4/registry (9fafbcfd): the renamed seats (issue 1966) 7c617bb085

Provision and provider users take the provisioner's client; mesh-vault writes its findings through the conditions client (K-4..K-20, K-21 part; ADR 0634 §4; issue 1966) e8f93d5db0

Build, image, node, distribution and gitea take parts, buildstep and composer from mesh-delivery-sdk; seatgen's client check allows a client's own packages and its -go-module has no root default; the build module writes its finding through the conditions client (K-4..K-20 part; ADR 0634 §3, §4, ADR 0635 §4; issue 1966) b2db504ab2

Seat fixtures of the build module's tests carry their domain (ADR 0636 §5; issue 1966) 12b5c62304

The gate accepts the bridge's published forgemsg by its series (K-22; ADR 0634 §3, F3; issue 1966) 500154b865

gofmt of the files the hand-acts and conditions clients touched (issue 1966) 631f689f40

messenger: the ask-events entry's grant is the config value readers, as the events template takes it (K-24; ADR 0595; issue 1966) 60c66a3641

Seat fixtures of the claude-code and node-build tests carry their domain (ADR 0636 §5; issue 1966) 059fa5d4c4

The seat-list case waits for the reader's change instead of racing it; the node build's generated-package fixture carries its domain (issue 1966) 75cf83a656

seatgen: a client may import its contracts module's domain SDK (-domain-sdk; ADR 0634 §2; issue 1966) 9017edaf9e

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

claude-code: the again tool answers every operator's session and the healers; the seat-list case waits on the reader's Ready (ADR 0636 §7; issue 1966) c7ae96414f

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Readings and last good nights are the owners' published packages: disk-load's diskreadings and restic's backupgood, taken by series (B5b Left 3; ADR 0634 §4; issue 1966) e0eecbf3c8

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Findings are written and read in the seat mesh-conditions' layout: gitea, distribution, mesh-config, artifact-store-tools (B5b Left 5 part; ADR 0634 §4; issue 1966) b4dbda287b

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

The key keeper is ssh-client's published package sshkeys, taken by series by the three build modules; Said, KindBuilder, ExecRunnerClean from the Delivery sdk (B5d; ADR 0634 §1; issue 1966) c46009709d

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

seatgen writes a date-time as time.Time (a string in TypeScript), by value when required, by pointer when optional (B5d; ADR 0634; issue 1966) f61760f738

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

messenger, desk-channel, telegram, mesh-config take the channel benches' shapes from the operator-channel contract and its client (B5d; ADR 0634 §1, §6; issue 1966) 1d3ed6e78e

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-watcher names the self-check seat and its event from the seat's contract (B5d; ADR 0634 §1; issue 1966) 0ae418b43c

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-self-check: a run is the seat's checked event and a finding the seat mesh-conditions' value (B5d; ADR 0634 §1, §4; issue 1966) 97cc9b1ccf

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-healers: attempts, would, the healer-acted event and its findings are the seats' contract types (B5d; ADR 0634 §1, §4; issue 1966) 916bdde6a4

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

The writers of findings give the seat's time.Time (artifact-store-tools, distribution, gitea, mesh-config; B5d; ADR 0634; issue 1966) 4104143d16

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

distribution: its wake events are heard through the events hub, the module's one durable pull (B5d; B9d hub; issue 1966) 62bb25675d

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-config and nats consume the seat mesh-catalog's manifest-accepted, not the indexer's (B7d finding 1; ADR 0634 §1; issue 1966) b0f2441dd3

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-conditions keeper works over the seat's types: conditions, events and findings in the seat's layout, the old layout still read (B5e; ADR 0634 §1, §4; issue 1966) c39c3fdddd

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

minio reads who holds a seat through the placement client and the seat's state name, not go/delivery (B5e; ADR 0634 §1, §4; issue 1966) 45f4edbb36

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

nfs-server reads who holds a seat through the placement client and the seat's state name, not go/delivery (B5e; ADR 0634 §1, §4; issue 1966) 1d20825d69

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-watcher reads who holds a seat through the placement client and the seat's state name, not go/delivery (B5e; ADR 0634 §1, §4; issue 1966) 4923e44cf7

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

gitea asks the builder through the builder's published package builds, taken by its series (B5e; ADR 0634 §1, §4; issue 1966) 72dd8a719e

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

distribution asks the builder through the builder's published package builds, taken by its series (B5e; ADR 0634 §1, §4; issue 1966) a636dc3df3

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

claude-code reads placements, modules and manifests through the placement and catalog clients and the seats' state names (B5e; ADR 0634 §1, §4; issue 1966) a609ebc8e5

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

disk-load reads the node-engine's data state through the node-engine client and the seat's types (B5e; ADR 0634 §1, §4; issue 1966) 8fabdc4b91

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-provisioner-client reads the placement and the node's have through the placement and node-engine clients and the seats' types (B5e; ADR 0634 §1, §4; issue 1966) 597910c291

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

pacman reads a node's declared packages in the node-engine seat's layout, not go/delivery's (B5e; ADR 0634 §1, §4; issue 1966) 0f1d4c5e0f

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-conditions' hold reads wants, haves and placement through the seats' types and clients; the motion is mesh-sdk's go/motion (B5e; ADR 0634 §1, §4; issue 1966) 42387b660c

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

artifact-store-tools reads the builder's builds through its published package, the wants, haves and manifests through the seats' types and clients; the sweep seat's two bodies are the holder's own (B5f; ADR 0634 §1, §4; issue 1966) d4abb82b07

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-provisioner reads the placement, wants, haves, compositions, modules and manifests through the seats' types and clients and the builder's builds through its published package (B5f; ADR 0634 §1, §4; issue 1966) 6e4de66df7

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-config reads the placement, wants, haves, modules, manifests, predefined list and config candidates through the seats' types and clients, the builder's builds through its published package; the kept value is its own layout (B5f; ADR 0634 §1, §4; issue 1966) 84a7639042

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-delivery-gate reads and writes its values through the seat's types and meshdeliveryclient's codecs, the view's reach, the checker's verdict key; the holder names are its own constants; no go/delivery left (B5h; ADR 0634 §1, §2; issue 1966) 72f22b11ec

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-build-go runs its asks on the Delivery SDK's neutral build ask, step and answer forms instead of mesh-sdk's go/delivery: the Puller takes buildstep.StepAsk (the read bridge is the form's own AskAs), the toolchains value is the build runner's own, the stale-step part is buildrun's, keys and values come from the seats and clients (B5i; ADR 0634 §1, §6; issue 1966) 8786bb3ba1

The part build-step lacks the check job's toolchain and types the check answer's fetch as the fetch arguments; both are kept in askrun_gaps.go until the part has them.

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-build-go: the imports the move added sit in the module group (B5i; issue 1966) cfde90906f

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-build-image runs its asks on the Delivery SDK's neutral build ask, step and answer forms instead of mesh-sdk's go/delivery; its state names are its seat's (B5i; ADR 0634 §1, §6; issue 1966) aa29d8feb8

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

artifact-store-tools answers held and says artifacts-gone in the seat's generated types (B5g, B2f; ADR 0571 §2, ADR 0634 §1; issue 1966) 078f1e5fa9

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-vault reads signers, node logins, tokens, placement, haves and the handing through the seats' types and clients; its holder names are its own constants (B5g; ADR 0634 §1, §4; issue 1966) d8b4bb5487

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-build-node runs its asks on the Delivery SDK's neutral build ask, step and answer forms instead of mesh-sdk's go/delivery; its state names are its seat's; the three build modules require the contracts they import (B5i; ADR 0634 §1, §6; issue 1966) 941959198e

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-config keeps held values in the seat's typed state kept, through the config client's key and codec (B5g, B2f; ADR 0565 §5, ADR 0634 §4; issue 1966) 1e974ce9e8

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-build-go: a seat's ask type runs as the neutral ask (AskOf, the step description, the fetch arguments) (B5i; issue 1966) 8f1581f34a

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-self-check reads the placement, seat holders, presence and health reports through the seats' types and clients; RootContract is the node-engine client's (B5g; ADR 0634 §1, §4; issue 1966) b8642c8b27

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

distribution steps its windows with mesh-delivery's sdk window package (B5g; ADR 0634 §1; issue 1966) 85999b7efd

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-build-go, -image and -node declare the builder's published package builds, which they now read their forms and keys through (B5i; ADR 0634 §1, §4; issue 1966) 634abead7d

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

gitea's whole test names the builder's versions address from the builder's builds package (B5g; ADR 0634 §1; issue 1966) df53a0c59e

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

nats reads manifest-accepted in the seat mesh-catalog's typed event and subject (B5g; ADR 0604, ADR 0634 §1; issue 1966) da560f90f8

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

checks/readable knows the owners' published constants (diskreadings, backupgood), not mesh-sdk's health ones the modules no longer name (B5g; ADR 0634 §4; issue 1966) ffa026f7d8

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Merge branch 'b5i/registry' into feat/registry-milestone 725ea4dba4

Merge branch 'b5/registry' into feat/registry-milestone 2685b9c749

mesh-build-go reads a check's toolchain from the job, answers with the Delivery SDK's CheckAnswer and typed FetchAnswer, and takes the toolchains value from the Delivery SDK; askrun_gaps.go and toolchain.go are gone (B5j; ADR 0584 clause 2, ADR 0634 section 6; issue 1966) ef81dc2712

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Merge branch 'b5i/registry' into feat/registry-milestone 36da33c818

seatgen writes a scalar typed [T, null] as *T in Go and T | null in TypeScript, so a declared zero survives with its type (ADR 0634; issue 1966) 75959dfa25

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-self-check reads a node's presence through the watchdogs client: a part's key is no node, a refused version is an error (ADR 0634 section 4; issue 1966) 73eb3bd11c

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

gitea and mesh-provisioner tests name the finding key and the findings state through the conditions client, not mesh-sdk's health (B5l; ADR 0634 §4; issue 1966) 8d3cffab9a

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-hand-acts and mesh-healers take the hand-act names and the act from the seat mesh-hand-acts and its client, not mesh-sdk's handacts (B5l; ADR 0634 §4; issue 1966) 6f2e27f59b

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-vault takes the internal name from mesh-wireguard's overlay package; the digest test no longer names mesh-sdk's buildstep, it takes the one package of its go list capture (B5l; ADR 0634 §4; issue 1966) 37cdf41088

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Merge branch 'b5/registry' into feat/registry-milestone bbcc95e66d

The nats keeper takes the removal grant from the grants part's published busgrants package and names it in packages (ADR 0634 §4; issue 1966) 1f3a293988

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-healers and mesh-self-check use the conditions client's Observation, Finding encoding and Says, not copies of their own (ADR 0634 §4; issue 1966) 2978dd5bed

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Builders, the artifact store's tools and distribution read and write the registry reference form through the registry clients' reference package, and name the sweep seat from its contract (ADR 0634 §4, S-8; issue 1966) 82d00aeb5e

catalogue's SeatReference, InRegistrySeat, Routed, Recorded, InArtifactStore and ArtifactSweepSeat leave mesh-sdk;
each package gives the two registry seats' names from their contracts (or the build step's copies of them).

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

nats and mesh-config: their event streams keep their history, the bus's predefined form of an event family, not an age-limited stream of their own (V1; ADR 0592 §6; issue 1966) a89aaf5644

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

artifact-store-tools: the sweep seats read the artifact store's form through the registry clients' forms too (S-8; issue 1966) 50acf7c1b4

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Merge branch 'b9i/registry' into feat/registry-milestone 8a8ad7713e

Require every mesh package of the group at the floor version (ADR 0573 sections 2 and 4, issue 1966) fb6bf361e4

A member's go.mod names the oldest version a module accepts; the check builds the group's own heads whatever it names. go.sum lines of mesh packages are left out.

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-mail: its go.mod test no longer reads a release out of the mesh-sdk version (the floor version, ADR 0573) 76e25aeee0

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Require the external packages a module's build reaches (issue 1966) 720efe6339

Versions are the ones the workspace already resolves, with their go.sum lines.

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

No client names a holder: each reader keeps its own constant, held to its manifest (ADR 0625 section 4, issue 1966) 78a1b5648b

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

The wanted accounts' layout leaves mesh-sdk for the grants part's package grantsstate (ADR 0634 section 1, ADR 0635 section 4, issue 1966) 29bcaf2be5

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

RemovedDeliveryVerbs leaves mesh-sdk for the seat mesh-delivery's client (ADR 0634 section 2, issue 1966) d630635229

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

No client names the holder of the seat mesh-provisioner: each reader keeps its own constant, held to its manifest (ADR 0625 section 4, issue 1966) 625d627794

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

The generated TypeScript parser of a key form names its own parameter and locals apart from the form's parts 2dc9e07e88

A key form such as <node>.<module>.<key> made the parser declare the parameter key and the part key together (TS2300 Duplicate identifier), which the per-package unit level now surfaces. The parameter and the locals are now _key, _rest and _at<n>; a test row with such a form fails on the old output. Issue 1966.

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

A section registered before the controller's terminal took over, or set there, is replaced or removed by a proposal the operator approves 657cb2b6c8

It could not be changed at all: propose refused a name set there and never replaced one. The ask now says plainly that
it replaces or removes the registered section and binds its text's digest; on Approve every node renders the new text
in the place of the old (a removal leaves it out); a replacement keeps the section's own heading form (issue 340 note 14).

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Every module follows the assigner's new places: the contracts' go/ path, mesh-wants in the delivery contracts, the state and tool assignments (issue 1966) c9d8950fd0

The legacy key form placement.<node> is gone from the client, so the readers and their fixtures lose it. The check noplacement
now also refuses the assigner's retired names in a manifest.

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Merge branch 'reg/claude-code-replace' into feat/registry-milestone d8227222a7

No checked-out library copy or comment says placement; the claude-code fixture and seat-list test pass at the merge (issue 1966) 6fb27549c9

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

A removal of an approved replacement names the registered text's fingerprint; the cases ADR 0642 cites are tests b3fc5b4580

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

A removal of a section removed already is refused: the case ADR 0642 names, as a test 80f992e367

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

A removal of a replacement names both fingerprints; the two-texts refusal names the places and the way out aa7c9e8924

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

A replacement is refused like any section, and an expired one changes nothing: two cases of ADR 0642 as tests 5aaafd917f

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Merge main (#617 , #618 ) into the registry milestone (issue 1966) 9c9c8a5631

mesh-build-go keeps trunk's zero-version resolution on the group's mesh-delivery sdk types; mesh-build-node keeps per-path packages with the builder's version and seat.

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Complete go.mod and go.sum for the externals each module builds with 35f005e53d

A check's tests run without network, so each module must carry every external module it builds with.

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Take tidy's go.mod whole: mesh requirements at the zero version only, unused ones dropped 50a66399f6

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

The composer test lays every repository beside the check as one workspace, as the group places the composer (issue 1966) e1aab8a892

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

The composer cache key covers the mesh package versions the ask names (issue 1921) 3b462cac31

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Merge the release fix/1921-zero-version-builds into the registry group: mesh-build-go keys a composer by the versions its compose ask names, in the group's typed fetch (issue 1921, issue 1966) 21eae590e0

The release read the versions from the untyped fetch's packages argument; the group's ComposeJob carries them in
buildstep.Fetch.Packages, so the group's form stands, and the release's check that the key differs for named versions
joins the existing test.

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

The composer test's workspace replaces every module it lays beside at the zero version (issue 1921, issue 1966) b02868f209

Go 1.27 looks the zero version of a workspace member up once a build loads the module graph; a versioned replace is the
form it takes with no lookup. workspaceOf writes the use list and those replaces, and a unit case loads such a graph with
GOPROXY=off. The only go.work the build path writes is this fixture's: the build module itself places requirements by
replaces in its modfile (pins.go), which are unversioned and not in a workspace, so they are untouched.

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-build-go takes a download again after a network failure, three times with a short wait, and says so (issue 1921, issue 1966) 77844477ec

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Merge the trunk (B0) into the registry group: one invokes list in mesh-build-go's manifest, the manifest asked of its owner and the checker's input asked of the seat (issue 1921, issue 1966) 8c67c224c2

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-build-go runs a program on the contract's run ask: the delivery domain SDK's job, key and bound, and the checker client's input ask (issue 1921, issue 1966) ba17347518

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

The unread-ask test and the ReadManifest note follow the delivery domain SDK's names and the ask to the indexer (issue 1921, issue 1966)

mesh/check failed (3): ace: the check of novox/mesh-catalog#612 failed: the unit level: the candidate's mes…; and 2 more | 7bf15a3dc4db on f2b332de73d6

mesh/delivery-group group feat/registry-milestone rejected: composed together: the checker's verdict at the members' heads (616c42434ac2): ace

Details

mesh/delivery closed: a newer head of the same pull request

Details

7bf15a3dc4

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

jschoubben force-pushed feat/registry-milestone from 125929dbc4 to 7bf15a3dc4

2026-11-07 02:54:16 +00:00

Compare

jschoubben added 1 commit 2026-11-07 03:50:42 +00:00

mesh-build-go requires the checker client it imports; a check refuses any module importing a mesh module its go.mod does not require (issue 1921, issue 1966)

mesh/check failed (4): unit: git.novox.be/novox/m…; and 3 more | 62fe1cc36d30 on f2b332de73d6 | judged by the candidate's mesh-build-go at e8757a23634d

mesh/delivery-group group feat/registry-milestone rejected: composed together: the checker's verdict at the members' heads (8a80e8e86858): unit: git.novox.be/n…

Details

mesh/delivery closed: a newer head of the same pull request

Details

62fe1cc36d

Quick fix: add the one require to mesh-build-go's go.mod (a minute). Thorough fix, chosen because the kind was met twice (a
local go.work resolving what go.mod never names) and hides itself from every local build: checks/requiredmodules reads every
go.mod and Go source (tests and tagged files included), finds the owning module by the longest module line among the
repositories beside, and fails in go test and the repository check before any push. The scan over the seven group heads
found this one instance.

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-tools commented

2026-11-07 03:51:12 +00:00

mesh-delivery: the operator's Approve of this pull request's merge into main is ended: a head it covers moved on: the heads it covers are no longer the group's heads. Nothing is merged on it. The operator is asked again once the checks pass at the new head.

jschoubben added 3 commits 2026-11-07 06:33:58 +00:00

Ignore the requiredmodules check's program, as the gitignored check asks of every main package (issue 1921) 81874e4748

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

The composer's test workspace lays what the composer requires, found in the group's trees and the catalogue's own (issue 1921) 91edd0bfdb

The group check of 8a80e8e86858 laid every Go module beside the check, and one of them (mesh-delivery's
mesh-grants, which now requires disk-load's diskreadings) required a catalogue module the catalogue under check
never lays beside itself: Go looked its zero version up and the composer did not build.

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

The composer's test workspace prefers the catalogue under check over its trunk laid beside (issue 1921)

mesh/check checking 422b22ab0e1a on 0f5201387891 | novox/mesh-catalog#612 in group feat/registry-milestone

mesh/delivery-group group feat/registry-milestone checking: 7 of 7 member(s) ready

Details

mesh/delivery closed: a newer head of the same pull request

Details

422b22ab0e

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

mesh-tools commented

2026-11-07 06:34:29 +00:00

mesh-delivery: the operator's Approve of this pull request's merge into main is ended: a head it covers moved on: the heads it covers are no longer the group's heads. Nothing is merged on it. The operator is asked again once the checks pass at the new head.

mesh-tools commented

2026-11-07 10:30:35 +00:00

mesh-delivery: not merged. The merge of this pull request into main at 422b22ab0e was declined: the operator, via telegram (user id verified), chose Decline. A new head asks the operator again.

mesh-delivery: **not merged.** The merge of this pull request into main at 422b22ab0e1a4f72bf120d9c473c1567b49269ca was declined: the operator, via telegram (user id verified), chose Decline. A new head asks the operator again.

jschoubben added 4 commits 2026-11-07 13:39:01 +00:00

The compose test downloads the composer's modules as the fetch step does, so no cache of this machine decides (issue 1921, milestone 47/G, the group check of c5d32074c945) ea25e4d589

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

The lab is marked moved: its forge files are templates filled from the forge-account binding, which the controller cannot answer (issue 1966) 962e8547b1

The check's fetch downloads the graph of a sibling Go module laid with the trees it requires; the compose test fetches nothing (issue 1921, milestone 47/G, the group check of c5d32074c945) 81287cb11d

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Merge origin/main: the trunk's fetch downloads each placed sibling's module graph (mesh-catalog #624 ) over the group's earlier version (issue 1921, milestone 47/G)

mesh/delivery-group group feat/registry-milestone checking: 7 of 7 member(s) ready

Details

mesh/delivery ready: it is merged on the Approve

Details

mesh/check checking 67d9ccb1d455 on 0f5201387891 | novox/mesh-catalog#612 in group feat/registry-milestone

67d9ccb1d4

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i

Some checks are pending

mesh/delivery-group group feat/registry-milestone checking: 7 of 7 member(s) ready

Details

mesh/delivery ready: it is merged on the Approve

Details

mesh/check checking 67d9ccb1d455 on 0f5201387891 | novox/mesh-catalog#612 in group feat/registry-milestone

You are not authorized to merge this pull request.

This pull request can be merged automatically.

View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.

git fetch -u origin feat/registry-milestone:feat/registry-milestone

git checkout feat/registry-milestone

Sign in to join this conversation.

2 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: novox/mesh-catalog#612