Replace native job/control polling with client-initiated WebSockets, durable
state and receipt reconciliation, independent verified document streams and
native administrator settings for transfer, connection and admission limits.
Preserve QQQ current-user authorization and device binding on every operation.
Validation: Maven verify passes 117 tests with zero Checkstyle violations;
OpenAPI generation/check and diff checks pass. Consumer CI pins protocol
67be986cb76d6b5d9a9f72cad054193b0d11b0d5. Installation is separate.
Co-Authored-By: Codex (GPT-6) <noreply@openai.com>
Signed-off-by: James Maes <james@kof22.com>
Reject bound device tokens on every legacy device route. Native proof verification
continues through standard QQQ sessions and current user permissions. Disable
legacy enrollment by default and explicitly in the native host; legacy-only demo
and fixtures opt in, preserving eligible unbound-device migration.
Validation: 110 Java tests pass with zero failures/errors/skips and zero Checkstyle.
Actual native-bind HTTP regressions reject bearer-only claim/content/queue/control;
legacy enrollment denial preserves the code for successful native enrollment.
Independent review found no remaining blocker. No schema or dependency changes.
Co-Authored-By: Codex (GPT-6) <noreply@openai.com>
Signed-off-by: James Maes <james@kof22.com>
Record verified RC20 behavior: OAuth2 caching defaults off and reruns supported
host authorization hooks; explicitly cached sessions require normal invalidation
or a qualified freshness policy. Table-based sessions reread user rows but do
not automatically interpret arbitrary enabled fields.
Validation: exact frozen source and a10-check synthetic reproduction. No product
code or provider/runtime configuration changes. Long-lived provider renewal and
configured native grants remain deployment qualifications.
Co-Authored-By: Codex (GPT-6) <noreply@openai.com>
Signed-off-by: James Maes <james@kof22.com>
Wrap standard extraction with explicit current-user enforcement after callbacks,
and authorize preview counts before the standard tenant-filtered count. Apply
the wrapper to all14 printing selection processes so trusted internal defaults
cannot skip table policies before Transform/Load checks.
Validation:109 Java tests pass with zero failures/errors/skips and zero
Checkstyle. New regressions exercise count/query denial, callback weakening,
selected/filter tenant isolation and process registrations. Native API classes
remain byte-identical to the23-check real client/server interop artifact.
Co-Authored-By: Codex (GPT-6) <noreply@openai.com>
Signed-off-by: James Maes <james@kof22.com>
Provide durable jobs, enrollment, tenant-scoped printer selectors and record actions, operational home widgets, protocol integration, and release automation.
Co-authored-by: Codex (GPT-6) <noreply@openai.com>
Signed-off-by: James Maes <james@kof22.com>
james.maesandCodex