Milestones
List view
Unversioned bucket: accepted ideas and post-v21 product features with no release thesis yet (explain/certify tooling, idea cards, per-writer encryption). Not release-gating.
No due date•0/23 issues closedUnversioned bucket: test-suite health, ratchets and gates, docs process, benchmarks, and repository tooling. Not release-gating. Cards graduate into a versioned milestone when a release thesis needs them.
No due date•0/51 issues closedPatch thesis: preserve the published 16.x graph semantics across garbage collection and later replica joins. Confirm the v16.0.1 dead-owner property sweep and membership retirement hazards with deterministic witnesses, then ship the smallest compatible correction as v16.0.2 on maintenance/16.x. Must ship: node re-add, delayed edge-property merge, and stale/concurrent membership witnesses; retained causal evidence where retirement is unproven; release and downstream artifact verification. Excludes new APIs, schema upgrades, mainline forward ports, and a general safe tombstone-retirement protocol. An end-of-life decision requires explicit downstream confirmation and is not assumed.
No due date•0/1 issues closedThesis: Introduce authoritative canonical/enriched merge nouns, local-site contracts and witnessed causal-slice collapse with separate admission. Add lowering, agent projections, generated merge contracts and lawful unmerge so consumers can reason about plurality, obstruction and repair without textual conflict folklore. Version boundary: MAJOR: new admission/merge and reversal contracts must explicitly distinguish observer projection from shared canonical truth and reversible Witness from operational receipts. Must ship / required disposition: #459, #462, #481, #484, #567, #568, #569, #570, #571. Exit gates: #484 -> #567 -> #568 -> #570 and #569 -> #570; #569 -> #462 -> #459; lawful forward collapse and sufficient witnesses precede #481. Compiled helpers consume authoritative runtime nouns. Multi-writer, plurality, refused admission and local-reassembly witnesses must pass in Docker. Out of scope: Picking a winner by default, granting admission merely by constructing a derived lane, making every open theory question a blanket blocker, and assigning Continuum/XYPH product meaning to git-warp. Planning policy: Must-ship items and their prerequisites define the intended delivery commitment. May-slip and discovery cards are candidate homes. Before release-prep, move unfinished candidates/containers to a later home or remove their release target: the existing release guard requires ZERO open non-release issues in the target milestone and ZERO open prior-release issues. Do not bypass that gate. Discovery must settle its question before implementation activation. Tracking containers are not additional PRs. Same-release prerequisite PRs land first. Dates are unset. Registry verification and retrospective close before the next train activates. Candidate scope by commitment: must-ship: #459, #462, #481, #484, #567, #568, #569, #570, #571. may-slip: #402, #711. discovery: #214, #460, #467, #468, #475, #486, #566, #625, #707.
No due date•2/22 issues closedThesis: Publish real lock-step ORSet, kernel and adapter packages and a versioned Wesley target, while keeping root consumers working. Applications gain reusable implementation and generation boundaries instead of private workspace shells or copied renderer recipes. Version boundary: MINOR only if root exports, CLI, storage and supported types remain compatible. Any required public removal moves the affected work to a major release. Lower packages must be published and available before shipped root imports consume them. Must ship / required disposition: #116, #515, #516, #517, #519, #828. Exit gates: #519 -> #517 -> #516 -> #515 with one repository tag and synchronized public versions. Installed consumers, browser/JSR imports, type-only coverage, deterministic SDK regeneration and invalid-schema diagnostics must pass. Audit the actual exported CRDT and port contracts before activating extraction; promote necessary candidate debt into must-ship scope. Out of scope: Relative imports into unpublished workspace shells, unvalidated public CRDT operations, application semantics guessed by the compiler, and a new per-package tag scheme. Planning policy: Must-ship items and their prerequisites define the intended delivery commitment. May-slip and discovery cards are candidate homes. Before release-prep, move unfinished candidates/containers to a later home or remove their release target: the existing release guard requires ZERO open non-release issues in the target milestone and ZERO open prior-release issues. Do not bypass that gate. Discovery must settle its question before implementation activation. Tracking containers are not additional PRs. Same-release prerequisite PRs land first. Dates are unset. Registry verification and retrospective close before the next train activates. Candidate scope by commitment: must-ship: #116, #515, #516, #517, #519, #828. container: #123.
No due date•0/44 issues closedThesis: Replace repeated manual checks and duplicated test scaffolding with one structural conformance producer, a diff scorecard/hook, catalog-backed executable documentation, and typed controller fixtures. Maintainers gain reproducible evidence while the public runtime remains compatible. Version boundary: MINOR: additive tooling and supported engineering workflows; retain public runtime, CLI and storage contracts. Refactors must independently satisfy the repository coverage and architecture gates. Must ship / required disposition: #363, #405, #414, #427, #437, #438, #441, #444, #578, #669. Exit gates: Shared producer/consumer ownership follows #437 -> #405 -> #438 and #427 -> #414 -> #363. Docs runners use #669. Each executable issue produces one green, independently mergeable PR; all test/benchmark execution uses COPY-based Docker. Out of scope: Making every optional automation tool a prerequisite for every PR, moving private package shells into shipped imports, and adding speculative domain APIs merely to improve scores. Planning policy: Must-ship items and their prerequisites define the intended delivery commitment. May-slip and discovery cards are candidate homes. Before release-prep, move unfinished candidates/containers to a later home or remove their release target: the existing release guard requires ZERO open non-release issues in the target milestone and ZERO open prior-release issues. Do not bypass that gate. Discovery must settle its question before implementation activation. Tracking containers are not additional PRs. Same-release prerequisite PRs land first. Dates are unset. Registry verification and retrospective close before the next train activates. Candidate scope by commitment: must-ship: #363, #405, #414, #427, #437, #438, #441, #444, #578, #669. may-slip: #112, #117, #119, #120, #122, #127, #132, #133, #184, #201, #212, #213, #237, #262, #265, #276, #366, #390, #391, #393, #397, #404, #410, #411, #412, #415, #416, #417, #419, #425, #428, #439, #440, #442, #445, #451, #478, #500, #601, #602, #603, #619, #668, #671, #684, #690, #691, #803, #804, #841, #856. discovery: #121, #130, #161, #167, #193, #200, #215, #216, #249, #365, #371, #423, #426, #861. container: #261, #364, #580. V20 retrospective intake (available follow-ups; not an activated train): #954 removes the unused dangling Dockerfile alias; #955 provides opt-in selected-lane Git setup plus cross-repository attachment proof; #956 improves bounded registry visibility handling. Each is assigned to James Ross with priority:next. Existing #691 owns the observed global-ASAP/milestone consistency scenario.
No due date•23/23 issues closedThesis: v18.2.1 is a patch release for WARP-owned materialization cache behavior and release documentation/evidence. Must ship: merged state-cache materialization fixes, current-version metadata, changelog/signpost updates, and the technical teardown. Not included: new public API, storage-format migration, or XYPH dependency adoption.
No due dateThesis: Build on bounded retained roots to expose honest compaction receipts and improve deployment operations with graceful sync-server shutdown and OS-vault-backed trust signing-key provisioning. Retention behavior must preserve stale-replica safety and make guarantees inspectable. Version boundary: MINOR: additive operations and receipts with existing persisted interpretation and public behavior preserved. Encryption/trailer format changes remain in the v21.0 major bucket unless separately proved compatible. Must ship / required disposition: #179, #450, #474. Exit gates: #911 safety precedes #474 supported compaction receipts. Exercise stale merges and checkpoint round trips; graceful drain and timeout outcomes; supported key provisioning without plaintext secrets in artifacts. Out of scope: Treating a local applied frontier as global causal stability, silently discarding tombstones, and conflating trust signing keys with data encryption-key lifecycles. Planning policy: Must-ship items and their prerequisites define the intended delivery commitment. May-slip and discovery cards are candidate homes. Before release-prep, move unfinished candidates/containers to a later home or remove their release target: the existing release guard requires ZERO open non-release issues in the target milestone and ZERO open prior-release issues. Do not bypass that gate. Discovery must settle its question before implementation activation. Tracking containers are not additional PRs. Same-release prerequisite PRs land first. Dates are unset. Registry verification and retrospective close before the next train activates. Candidate scope by commitment: must-ship: #179, #450, #474. may-slip: #129, #180, #236, #401, #406, #487. discovery: #138, #139, #140, #141, #142, #452.
No due date•4/25 issues closedThesis: Replace production full-state replay/materialization contracts with indexed retained roots, bounded causal reduction and root commitments. Add typed recursive graph references with lawful ownership, cycle-safe traversal and retention; migrate real repositories through the same bounded paths. Large graphs and assets become process-memory-bounded rather than merely generator-shaped. Version boundary: MAJOR: retained substrate/commitment and migration boundaries change. Keep legacy decoding in migration scripts, preserve source authority until verified promotion, and provide an explicit upgrade/recovery contract. Must ship / required disposition: #192, #228, #248, #811, #814, #815, #816, #817, #819, #820, #821, #822, #823, #917. Exit gates: Respect #815 -> #816 -> #817 -> #248 -> #814; ownership #903 precedes reference/traversal/retention work. Complete #822 migration and #823 global heap/RSS proof including #917 public streaming ingestion. Reopen, history, receipts, cancellation, recursion and GC retention all need production-path witnesses. Out of scope: Observer-relative braid admission, unmerge, undocumented root compatibility aliases, and arbitrary graph-scale in-memory collectors. Planning policy: Must-ship items and their prerequisites define the intended delivery commitment. May-slip and discovery cards are candidate homes. Before release-prep, move unfinished candidates/containers to a later home or remove their release target: the existing release guard requires ZERO open non-release issues in the target milestone and ZERO open prior-release issues. Do not bypass that gate. Discovery must settle its question before implementation activation. Tracking containers are not additional PRs. Same-release prerequisite PRs land first. Dates are unset. Registry verification and retrospective close before the next train activates. Candidate scope by commitment: must-ship: #192, #228, #248, #811, #814, #815, #816, #817, #819, #820, #821, #822, #823, #917. may-slip: #137, #171, #186, #188, #190, #207, #211, #241, #243, #252, #422, #424, #431, #433, #454, #458, #470, #488, #528, #704, #812, #896, #897. discovery: #146, #149, #158, #164, #175, #185, #209, #223, #225, #233, #238, #239, #242, #246, #247, #375, #383, #388, #396, #399, #418, #430, #453, #455, #456, #457, #463, #464, #465, #476, #483. container: #224, #226, #232, #523, #565, #824, #905.
No due date•102/166 issues closedThesis: Make ordinary v20 Runtime applications reliable for observed writes, bounded tail reads, checkpoint maintenance, saved history, strands, and safe Git transfer. Preserve restart-stable occurrence relations, materialization-free provenance, and truthful query/admission explanations. Version boundary: MINOR. Preserve v20 signatures, stored identities, interpretation, and existing failure meanings. Add new validated operations where required. Record any breaking-contract conflict before activation; do not hide it within this minor release. Independent agy review on planning head 6e9c1bf9 returned REQUEST CHANGES. The full feedback is preserved in Reader under receipt dd2bd42e-da6f-4e63-b4e6-d51d7b8e02d9. docs/plans/v20.1.0/agy-review-resolution.md records accepted constraints and source-based qualifications. No approval of the revised head is claimed. Current graph version 2026-10-07.8 has 22 recorded prerequisites and 5 topological layers. The original scope is preserved; audit and bounded-parent repairs add required executable cards. Mainline merging is paused. Six of 36 required tasks have ready, unmerged PRs; this is preparation progress, not release closure. #965 -> #964 owns canonical saved-coordinate restoration. #985 -> #957 fixes the SHA-256 bootstrap prerequisite. Package gates #958, #982, and #983 remain blocked until compatible upstream releases are verified. Required existing commitments: #392, #395, #421, #473, #854. Required fixes (28 study tasks, the SHA-256 prerequisite, the locked audit, and bounded checkpoint parents): #957, #958, #959, #960, #961, #962, #963, #964, #965, #966, #967, #968, #969, #970, #971, #972, #973, #974, #975, #976, #977, #978, #979, #980, #981, #982, #983, #984, #985, #987, #990. Inventory: 47 cards; 36 required, 6 candidates, 5 discovery. Each executable issue maps to one coherent, independently mergeable PR. Tracking and dependency records are plans, not runtime proof. Source: FEEDBACK-git-warp.md, 2026-10-06, SHA-256 9b15209d51cd705059a9e6bfebf0f6681ff832924c1462b99ff3cfebde6bacae. Review baseline: v20.0.0 / main ceb58e656ec5bc85f0ae5991bf2a55599693bbb3. Root ROADMAP.md and docs/plans/v20.1.0/tasks/ contain the complete linked task plans. Each new issue has a house template and an LLM prompt. Exit gates: Public installed-package regressions; fresh Runtime and restart cadence; CRDT parity; saved-coordinate GC/reopen; no hidden whole-graph reads; honest refusal/cost evidence; safe prune-enabled transfer; strict packed NodeNext consumers; measured install and startup costs. All tests and benchmarks run in guarded COPY-based Docker. External gates: Consume published compatible git-cas and schema/package corrections where required. Do not bypass retention or force incompatible dependency overrides. Candidates: #126, #274, #394, #466, #485, #489. Discovery: #220, #231, #254, #472, #477. Out of scope: General v21 retained-root migration, generic v22 merge/unmerge, rewriting history, hidden materialization fallbacks, and weakening release guards. The selected public slice and common-basis preparation defects remain required here; broader follow-ons retain their current owners. Before release-prep, explicitly dispose of each unfinished candidate and discovery card. The release guard requires zero open non-release target issues, zero open prior-release issues, and zero repository-wide ASAP issues. Required fixes cannot slip merely to obtain a zero count. No calendar date is assigned. Normal review, registry verification, and retrospective gates remain in force. The earlier 2026-10-01 GitHub/Linear snapshot remains historical. This update changes GitHub and local plans only; it does not claim a new Linear reconciliation.
No due date•37/37 issues closedPatch bucket only. Inactive until a patch thesis and at least one must-ship maintenance fix exist.
No due date