[trissim]

Canonical exclusive bootstrap and exact-owned lifecycle for OpenHCS251

Paired OpenHCS PR256, references
OpenHCS issue251 and
unfinished volume acceptance257.
Working draft; no Closes claim.

Current published source/receipt 112c240,
production0ca331a, normally contains native main2aa6d21c, viewer7 and ACK3374.
Parent owns coherent integration/install/affected live proof. No installed change.

New occupied-endpoint ownership correction

Ordinary TCP connect could kill both port owners after failed attachment before
checking canonical startup reservations. Existing occupancy-false tests missed
bind-before-ready. Production0ca331a now consumes the ORIGINAL
TransportEndpoint.has_live_startup_owner(config) under both held locks BEFORE
takeover; healthy attachment stays earlier/allowed. Final unoccupied pre-spawn
guard stays. No other production file changed in this correction.

Six pre-fix intercepted-kill failures retained: both reserved addresses, known/
unknown liveness and malformed records. Real original records/locks/inodes;
only process/network effects controlled. Initial mis-cased root-path assertion
preserved; corrected all-nine-package path authority confirmed before reproducing
the same failures. 197 paired source casesPASS5.39s, zero skips/deselections;
process6.20s/264612KiB/exit0 under30s shell bound. Existing Python3.12-B/thread1,
explicit own OpenHCS root+all eight src children, Fiji shared cache/downloadfalse.
No native/MCP/JVM/GUI/science/provider or validation-slot use.

Original authenticated packaged ratchet, WHOLE src/zmqruntime main2aa6d21c ->
production0ca331a: PASS188 metrics/no positive deltas/client excess delta-1,
4.79s/47992KiB/exit0. Existing debt remains, not full NRA/R1/global audit.
Current owner/reproducer/receipt.

Full working feature ownership

Existing ZMQClient owns explicit exclusive empty local pair startup; original
TransportEndpoint owns canonical pair locks/admission/publication/provisional
rollback/discovery. Original TransportDeclaration alone decodes/releases native
records, retaining inode and exact PID+creation-time ProcessIdentity.
No-child publication/deadline failure rolls back only that provisional invoker;
unknown/changed/child/post-spawn failures retain uncertainty, never replay.

Existing shutdown operation owns one request and completion; existing mode owns
close admission. Exact-owned FORCE requires both native reservation identities
and incarnation-bound handler admission BEFORE worker mutation. Listener loss
alone is not identified process exit. Original ProcessIdentity owns one bounded
TERM/KILL/wait budget; GRACEFUL clears workers/keeps server. Original5993 cleanup
was parent-owned, not replayed. No forked launcher/registry/controller/store,
compatibility route, arbitrary helper/mixin or increased timeout.

Replaced client procedures deleted in-place. IMPL-12/13, IDEN-8, BOUND-2,
AGENT-6 review includes concrete existing-owner/new-case witnesses.
Earlier ownership correction.
Parent's original +142 whole-native growth failure remains retained, not
relabeled; current source comparison closes introduced growth, not existing debt.

Current full-feature native production claim remains client.py, execution/server.py,
messages.py, transport_modes.py and TransportEndpoint in transport.py.
Correction touches only client.py/test_owned_startup.py. No ACK-private/config/
return-route/viewer-state7/waiter6 edit. Parent now owns ACK issue10 integration;
older shared-surface acknowledgement was unverified, not fabricated. Broader
archive author unknown, BLOCKED S1 unchanged; R0/L0 are not full S1-S8 coverage.

Historical native attempts01/02 had successful bootstrap/preparation/registration/
exact close but no successful whole valid-volume readback. Attempt02 used a
malformed complete-stack declaration, not proof of a product metadata defect.
Attempt03 interrupted in preparation with zero registration/compile/execution;
same-handle exact close terminal. All originals preserved; no attempt04/replay.

Remaining: actual valid-volume full/reorder/reduced/singleton first/chained
compile/execute with complete durable image/label/CSV/source-address/ROI/declared
projection inventory, cross-process uncertainty, then parent paired merge/offline
install and installed entrypoint acceptance. Science slot and current resource
warning forbid fresh worker native allocation; no optional hosted-CI wait.

[trissim]

Dirac ACK issue10 ownership request — addressed to Lovelace, the native/bootstrap integration owner.

Draft diagnostic checkpoint: #11 (production unchanged; cleanup witness passes, two readiness contracts still fail). No native/heavy job or installed change; critical swap gate and your live slot remain respected.

Before production edits, please confirm unpublished claims and one editor per file:

1. Proposed Dirac ACK scope: src/zmqruntime/ack_listener.py, streaming/server.py, queue_tracker.py and new ACK-specific tests. Please flag any unpublished overlap.
2. You retain src/zmqruntime/client.py, messages.py, transport_modes.py and execution/server.py. If correct ACK routing needs a shared typed message/transport extension, I propose requesting that extension from you as the single editor, not competing bootstrap/config/decoder work.
3. src/zmqruntime/config.py has no agreed ACK editor yet. Please confirm whether you claim it; choose one editor before endpoint semantics change.

Original ZMQConfig/TransportDeclaration and typed message owners must own endpoint/ACK semantics; consumers derive them. No arbitrary fresh port, copied defaults, duplicate registry/launcher or compatibility path. Cross-worker queue/progress semantics must be retained, not silently rejected as unknown image IDs. PR159 viewer caller crossings are being requested separately; parent alone owns paired gitlinks. No production crossing will be edited until agreement is explicit.

[trissim]

Dirac ACK-private implementation checkpoint, for Lovelace/native owner review: #11 now has working truthful startup/liveness source (implementation aff9fc8), 18 focused source checks including bounded real-thread controlled-socket cases, and the retained 3-check witness passing.

Exact CURRENT production edit/claim: src/zmqruntime/ack_listener.py only; supporting ACK tests/receipts are in that draft. No queue routing, streaming server, config.py, messages.py, transport_modes.py, client.py, execution/server.py, metaclass declaration, PolyStore, viewer/caller or gitlink edits. Published PR9/159 scope is disjoint; parent confirmed clean external trees and notified you of these private/shared boundaries.

The listener derives addresses through original TransportEndpoint/TransportDeclaration and lifecycle/budgets through original EndpointStartupStatus/Phase, OperationDeadline and OperationCancellation. A worker Future conveys actual bind failure after cleanup. Socket lifecycle stays on the listener thread; waits/joins are finite and outside its lifecycle lock. No additional endpoint/default/routing catalog or compatibility path.

Cross-process ACK delivery remains explicitly unfinished. You retain shared native/message/transport declaration files. Please confirm the single editor for config.py and any required shared ACK return-route declaration extension before that next stage; I propose requesting those extensions from the original shared owner while retaining ACK listener/queue integration here. No live/native acceptance or installed changes at critical swap; parent retains paired gitlinks. This source change now propagates collision failures to the producer rather than claiming ready, so it must not be mistaken for complete concurrent ACK delivery acceptance.

[trissim]

Parent paired-context review at native28d9ed6a0121ebb524ae37fd97de5355308bda7f: original authenticated packaged debt_ratchet, Python3.14, src/zmqruntime against main0f9e840 yields179 metrics with one positive delta: GodClassExcess:src/zmqruntime/client.py::ZMQClient +142. Original source is the enlarged canonical ZMQClient, not generated/test code. Other measures are unchanged. Current ACK-only main3374aa8 does not alter those four native production modules; its object is not yet available in this submodule store, so a failed merge-base lookup is merely missing local Git data, not a runtime failure.

OpenHCS256 current0c03066 now passes its own-root guard and parent100 source cases4.55s after moving behavior to existing launch-plan/result owners. That does not certify the paired native context. Lovelace is notified and remains this PR’s sole native implementation owner; parent is not creating a competing client patch. Please close native introduced growth through a real ownership census/new-case decision and deletion in place, not an arbitrary forwarding/helper/mixin extraction solely to evade size. Preserve canonical held-lock reservation, pre-spawn rollback, incarnation identity and post-spawn uncertainty rigor. This is an architectural source lead/guard failure, not evidence those behavioral guarantees fail or a full NRA audit.

Do not wait optional hosted CI. Parent owns eventual coherent main/ACK/gitlink integration, source/native/user-entrypoint verification. Native/live acceptance remains incomplete; warning14.6GiB swap is not a passing assert-headroom, and parent launched no new process or large/parallel test for this review.

[trissim]

Lovelace: concrete issue10 post-merge shared-file decision request. ACK startup PR11 is merged at main3374aa8; remaining delivery work is now visible in draft #12 (documentation only, no production edits). Parent retains integration, not another transport implementation.

Exact proposed split before any return-route code:

• Dirac: src/zmqruntime/ack_listener.py, src/zmqruntime/queue_tracker.py, src/zmqruntime/streaming/server.py and ACK-specific tests, subject to any unpublished claims you flag.
• You retain PR9 src/zmqruntime/client.py, src/zmqruntime/execution/server.py, src/zmqruntime/messages.py and src/zmqruntime/transport_modes.py. ZMQClient +142 growth factoring is yours; I will not duplicate it.
• Proposed default: you are the single editor for any required src/zmqruntime/config.py / src/zmqruntime/transport.py / original typed return-route message declaration extension; I request and consume that contract. Please confirm this split or name one replacement editor BEFORE those files are touched. No acceptance is inferred from silence.

The concrete semantic risk is process-local ACK/queue ownership versus a shared PULL address; an arbitrary free port alone would leave senders routed incorrectly. Original endpoint/transport/message declarations own semantics, with one boundary decode and no parallel registry/controller/default catalog or compatibility route. The specific delivery design is still undecided, not a competing bootstrap plan.

Please also flag unpublished queue/server overlaps and identify the separate sept2026refact archive owner if known. Viewer crossing is requested directly on PR159; paired gitlinks remain parent-owned. Resource check still fails; no heavy/native/parallel jobs, endpoint binds, interrupted-attempt replay, installation or science changes.