43-EFT.WP.Data.DatasetCards v1.0 | Chapter 14 Release & Versioning

Home ／ Docs-Technical WhitePaper ／ 43-EFT.WP.Data.DatasetCards v1.0

Chapter 14 Release & Versioning

I. Chapter Purpose & Scope

II. Terminology & Dependencies

• Terminology source: EFT.WP.Core.Terms v1.0; this chapter only adds release/version-specific fields.
• Dependent volumes: Data contract/export: Core.DataSpec v1.0; metrology/dimensional checks: Core.Metrology v1.0; splits & distribution: Chapter 11; quality & baselines: Chapter 12; privacy & compliance: Chapter 13; citation/version posture: Citation & Cross-Reference Spec v0.1.

III. Fields & Structure (Normative)

release:

channel: "stable" # alpha | beta | rc | stable | yanked

version: "v1.0.0" # semantic version: vMAJOR.MINOR.PATCH

date: "2025-09-20"

tag: "eift-foo-v1.0.0" # git/object-store tag

compatibility:

baseline: "v1.*" # public stability line

backwards: "minor" # none|patch|minor

artifacts: # must mirror export_manifest

- {path:"dataset_card.yaml", sha256:"..."}

- {path:"splits/train.index", sha256:"..."}

- {path:"packages/train-000.tgz", sha256:"..."}

checks:

schema_ok: true

lint_ok: true

check_dim_ok: true

quality_gates_ok: true

notices: # public notices (withdrawal/correction/limits)

type: "release" # release | correction | withdrawal

summary: "initial stable"

see:

- "EFT.WP.Core.DataSpec v1.0:EXPORT"

- "EFT.WP.Core.Metrology v1.0:check_dim"

versioning:

scheme: "semver"

stability_line: "v1.*" # recommended public citation

policy:

major: "breaking" # bump MAJOR only for breaking changes

minor: "additive" # backward-compatible additions/optionals

patch: "fix" # fixes/docs; no semantic change

deprecation:

announce_in: ["dataset_card.yaml", "export_manifest"]

grace_cycles: 1 # keep at least one MINOR cycle

withdrawal:

allowed: true

reasons: ["legal","safety","integrity","privacy"]

mirror_tombstone: true # mirrors retain tombstone & notice

(Fields release.* and versioning.* synchronize card and portal; all deliverables verifiable via export_manifest.)

IV. Semantic Versioning & Stability Line

1. Format: vMAJOR.MINOR.PATCH (e.g., v1.2.3).
2. Bump rules:
   • MAJOR: breaking change (required-field semantics changed/removed, split-index format change).
   • MINOR: backward-compatible additions (new optional fields, added mappings, examples/shards).
   • PATCH: fixes and minor corrections (text/notes/check scripts) without semantic/structural change.
3. Stability line: Public citations should target v1.*; papers/benchmarks are advised to pin to a minor (e.g., v1.2.*) for reproducibility.

V. Compatibility Matrix & Deprecation

1. Compatibility matrix:
   • Schema: minor+patch compatible; major may break.
   • Metrics & evaluation protocol: additions are non-breaking; removals/semantic shifts require major.
   • Split indices: immutable once frozen; any change mandates major and a migration script.
2. Deprecation: Announce deprecated items and replacement paths in dataset_card.yaml and export_manifest.references[]; allow at least grace_cycles >= 1.

VI. Withdrawal & Correction

• Withdrawal: Triggered by legal/safety/integrity/privacy risks; set release.channel="yanked"; mirrors retain tombstone and link to public notice.
• Correction: Minor fixes via PATCH; if comparability is impacted, set notices.type="correction" and mark scope in the quality report.

VII. Pre-Release Blocking Checks

• schema_ok, lint_ok, check_dim_ok, quality_gates_ok are all true;
• splits are frozen and consistent with sampling/cleaning audits (Ch. 6–7);
• Privacy/ethics and regional compliance validated and reflected in distribution.regional_compliance[].

VIII. Export-Manifest Crosswalk (Normative)

export_manifest:

version: "v1.0.0"

artifacts:

- {path:"dataset_card.yaml", sha256:"..."}

- {path:"splits/train.index", sha256:"..."}

- {path:"packages/train-000.tgz", sha256:"..."}

- {path:"quality/summary.csv", sha256:"..."}

references:

- "EFT.WP.Core.DataSpec v1.0:EXPORT"

- "EFT.WP.Core.Metrology v1.0:check_dim"

notices:

- {type:"release", summary:"initial stable", date:"2025-09-20"}

(All artifacts must be verifiable; references carry Volume+Version+Anchor; no shortcodes/aliases.)

IX. Channels & Mirrors

• Channels: alpha (internal), beta (external feedback), rc (candidate), stable (public), yanked (withdrawn).
• Mirrors: Provide ≥2 mirrors with rate/region limits; on withdrawal, all mirrors display tombstone and link to the notice anchor.

X. Machine-Readable Fragment (Drop-in)

release:

channel: "rc"

version: "v1.1.0-rc1"

date: "2025-10-01"

tag: "eift-foo-v1.1.0-rc1"

compatibility: {baseline:"v1.*", backwards:"minor"}

artifacts:

- {path:"dataset_card.yaml", sha256:"..."}

- {path:"packages/train-000.tgz", sha256:"..."}

checks: {schema_ok:true, lint_ok:true, check_dim_ok:true, quality_gates_ok:true}

notices: {type:"release", summary:"release candidate for v1.1.0"}

versioning:

scheme: "semver"

stability_line: "v1.*"

policy: {major:"breaking", minor:"additive", patch:"fix"}

deprecation: {announce_in:["dataset_card.yaml","export_manifest"], grace_cycles:1}

(Consistent with export_manifest, Chapters 11–13 constraints, and metrology/uncertainty posture.)

XI. Chapter Compliance Checklist

• Version uses semantic format; stability line is explicit; public materials cite v1.* or pin to a minor.
• Compatibility, deprecation, and withdrawal policies are announced and traceable in the card and export manifest.
• All pre-release checks passed; splits/quality/privacy and regional compliance are consistent.
• Export manifest is complete and verifiable; all references carry “Volume+Version+Anchor,” with no shortcodes/aliases.