Repository navigation
Roadmap Mathematical Ownership
Chris Dare edited this page Nov 15, 2026
·
1 revision
schema: roadmap/1
slug: mathematical-ownership
project: derived-alg-geo-lean
title: Mathematical ownership and abstraction boundaries
status: active
phase: sequenced
brief: 'Structural companion to charge-abstraction CA1–CA3 (milestones 49–51), based on the 2026-09-13
repository-wide ownership review at 9edfb9d6. It preserves the existing single charge kernel, separates
foundations from comparisons/applications, and repairs the non-charge owners without duplicating SRF1,
DG3, DT1 or SF8 theorem work. Source review: https://github.com/chris-dare-dev/derived-alg-geo-lean/blob/965d62fc66afbbe998482ef0912cacc265a01c5a/docs/reviews/2026-09-13-mathematical-ownership-review.md.
Execution plan: https://github.com/chris-dare-dev/derived-alg-geo-lean/blob/965d62fc66afbbe998482ef0912cacc265a01c5a/docs/reviews/2026-09-13-mathematical-ownership-plan.md.'
goal:
objective: Place independently reusable mathematics at its proper owner, retain canonical roots and
explicit specialization/realization maps, and prevent foundations importing their applications.
key_results:
- All fifteen review findings have an implemented or explicitly reasoned disposition, linked to source
and implementing PRs.
- Charge construction precedes walls; neutral quadratic/lattice algebra and geometric Mukai interpretation
have distinct owners.
- Linear Serre/Yoneda, abelian stability, dg H0, derived operations, perfectness, GL cover and planar
geometry have independently importable foundations.
- CA1–CA3 retain their mathematical ownership and conditional root decisions; no formula, pairing or
carrier is duplicated by moving a file.
- Tracker, roadmap, imports, umbrellas, audits, source-owner checks and prescribed runner CI agree.
wont:
- No source refactor is implemented by the planning PR; proposed destinations are settled in MO1.01
before cutover.
- No global rank ordering of Mathlib subjects, no repository/library split, and no empty speculative
abstraction carriers.
- No automatic equivalence of numerical models with geometric realizations, kernel negativity with full
support, ordinary H0 presentation with exact enhancement, or arbitrary loci with codimension-one manifolds.
- No hidden strengthening of finite-Tor, Noetherian, HN, integrality, signature or exactness hypotheses.
- No completion claim for unrelated mathematics in CA, SRF1, DG3, DT1 or SF8.
generated_by:
agent: Codex
at: '2026-09-13'
generations:
- at: '2026-09-13'
note: Created at the user’s request as a companion milestone with milestone-direct tasks. New native
blockers are only on MO1 tasks; existing CA membership, status and native blockers are preserved.
The new milestone is first claimed by this roadmap file, avoiding new unreferenced issues on existing
roadmap-owned milestones before the planning PR merges.
items:
- id: mathematical-ownership-m1
gh_milestone: 52
kind: milestone
title: MO1 — Mathematical ownership and abstraction boundaries
status: planned
lane: now
acceptance:
- MO1.01 records canonical owners and reconciles pending path contracts; no new carrier bypasses the
root-review rule.
- MO1.02–MO1.14 implement or explicitly adjudicate findings 01–15 with their existing mathematical boundaries
intact.
- MO1.15 records focused import evidence, comparison preservation and required runner verification;
all tracker/roadmap references agree.
- id: mathematical-ownership-t1
gh_issue: 1312
kind: task
parent: mathematical-ownership-m1
title: MO1.01 — Agree canonical owners and a coordinated cutover map
status: done
progress: 'Two parts. PR #1327 codified the policy: CLAUDE.md, AGENTS.md and the architecture
documents distinguish ownership, imports and specialization, with subject-specific boundaries.
PR #1328 supplied the three things that left open — the declaration-level cutover map for all
fifteen findings in docs/architecture/cutover-ledger.md, the finer gate change (rule 3 named by
subcomponent instead of by blanket subtree, with regression fixtures, taking the guarded
stability-neutral count from 258 to 302 of 381), and the per-issue path reconciliation. Neither
part moves a Lean declaration; the source repairs are MO1.02–MO1.14.'
summary: Record the definition owner, neutral core, application adapter and comparison owner for each
confirmed finding before moving source. Amend the placement policy where it currently conflates a
new subject with an extension of an existing Mathlib API. Coordinates CA1–CA3 (#1222–#1230), SRF1,
DG3, DT1 and SF8. It establishes a shared placement contract; it does not block existing mathematical
proofs on completion of the entire MO1 milestone.
priority: must
size: M
tags:
- ownership
- architecture
- review-14
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:governance
- type:docs
- sev:medium
- ready
acceptance:
- Publish a declaration-level old-to-new owner map in docs/architecture/cutover-ledger.md. Consult the
pinned Mathlib definition site first; candidate paths in the review are proposals, not existing upstream
APIs.
- Document separately subject ownership, imports and specialization maps. Retain one DerivedAlgGeo library,
the geometry firewall, Development as a leaf, and generic derived categories and DGCategory beside
their actual Mathlib carriers.
- 'Settle the shared CA path contract: neutral paired complex functionals in linear algebra; stability
charge families upstream of Walls; geometric Chern/Todd realizations under AlgebraicGeometry. Preserve
the single roots and names required by #1223 and #1230.'
- Name component-specific forbidden import edges and legitimate core-umbrella exceptions; narrow the
blanket Numerical stability exemption by subcomponent instead of imposing a total order on Mathlib
subjects.
- Update AGENTS.md, CONTRIBUTING.md, placement.md, abstraction-tree.md and the cutover ledger consistently
where affected. Reconcile the path-specific clauses of linked open issues with this map without changing
their mathematical acceptance.
- Every proposed new carrier has the required independent-consumer justification or is replaced by a
theorem/abbreviation. No empty speculative hierarchy or mass rename is authorized by this decision
alone.
- id: mathematical-ownership-t2
gh_issue: 1313
kind: task
parent: mathematical-ownership-m1
title: MO1.02 — Make charge construction upstream of walls and separate neutral pairing lemmas
status: done
progress: 'Moved the unchanged ChargeFamily and Exp.ofMoments roots to StabilityCondition/CentralCharge,
split surface and threefold construction from their wall equations, moved divisorial charge data and
supplied Todd corrections upstream, extracted neutral paired-functional, continuity, bounds and Hodge
owners under LinearAlgebra, and moved full divisorial support under Support. Rule 8 pins the owners and
transitive import boundaries with forbidden fixtures. Issue #1230 was withdrawn on 2026-09-30 as a clause 2a root review, so this cutover retains PeriodDomain.centralCharge and Mukai.expCharge with their existing
bridges; no pairCharge or graded root was invented.'
summary: Split the linear-algebra functional/kernel/bounds API from its stability interpretation, and
move the existing ChargeFamily and exponential/divisorial constructors to their agreed owners upstream
of Walls. CA1 owns formulas and corrections (#1222–#1225); CA2 owns rotation/tilt (#1226–#1228); CA3
owns the pairing bridge and conditional root (#1229–#1230). This issue owns extraction and import
direction only.
priority: must
size: L
tags:
- ownership
- architecture
- review-01
- review-04
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:stability
- area:algebra
- type:refactor
- sev:high
acceptance:
- 'Reuse ChargeFamily and Exp.ofMoments from #1219–#1221; all existing surface, threefold, slope and
divisorial comparisons remain available. No second charge carrier or polynomial is introduced.'
- Extract arbitrary quadratic continuity/coercivity/level-set bounds from WallFiniteness; retain genuinely
quadratic statements under QuadraticForm. General paired complex-functional algebra is independent
of stability, schemes and wall arrangements.
- Place stability-facing families and support adapters under the agreed CentralCharge/Support owners.
Split divisorial charge data, Chern coordinates, Hodge input and Todd corrections by actual subject;
wall equations import the constructors.
- Keep the primitive abstract charge additive (a map to ℂ); Chern character and κ are geometric constructors,
not mandatory fields of every charge. State both semistable-class nonnegativity and kernel negativity
when claiming quadratic support.
- Update the divisorial-root assertions in the layering/single-owner gates with the move. Demonstrate
imports of neutral pairing, charge-family core and Exp kernel do not reach their wall or geometric
consumers.
- '#1230 withdrew pairCharge on 2026-09-30: retain the existing functional plus its bridges. No placement split
waits on a graded structure.'
- 'Keep the existing tilt-dependent rotation comparison downstream of the neutral ChargeFamily core;
accommodate #1226–#1228 without importing tilting into it.'
blocked_by:
- 1312
- id: mathematical-ownership-t3
gh_issue: 1314
kind: task
parent: mathematical-ownership-m1
title: MO1.03 — Separate positive frames, positive planes and the different wall loci
status: done
progress: 'Separated the neutral positive-plane and ordered-positive-frame carriers and added
explicit framePlane and forgetPositiveFrame maps. Orthogonality arrangements and finiteness remain
in neutral QuadraticForm modules; generic charge-zero, determinant-alignment, positive-ray,
spherical nonpositive-ray and stability-space charge-zero loci now have distinct names and owners,
with comparison theorems carrying the required nonzero and sign data. Rule 10 pins those owners,
forbids a common codimension-one parent and retires the compressed source paths. Existing historical
spellings remain executable-only for immutable review restatement; no actual destabilization theorem
or geometric period-domain identification was invented.'
summary: 'Give the current period-domain and wall APIs names and owners matching what they actually
define, with explicit maps between frame, plane, numerical-locus and stability interpretations. Consumes
the ownership split in MO1.02. Relates to CA2 wall invariance, but does not duplicate #1226 or assert
phase invariance.'
priority: must
size: M
tags:
- ownership
- architecture
- review-02
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:stability
- area:algebra
- type:refactor
- sev:high
- ready
acceptance:
- Distinguish the current unoriented positive two-plane carrier from ordered positive frames. Provide
the required forgetful/comparison maps; do not identify a complex-vector domain with a submodule carrier
by a rename.
- Retain general positivity, signature, orthogonality arrangements and their finiteness lemmas in neutral
linear algebra. Place actual geometric period-domain identifications with the geometric realization.
- Document and separate Z(δ)=0, determinant/alignment loci, signed ray loci and actual destabilization
walls. Charge-zero is generically real codimension two; alignment is generically codimension one only
with appropriate nondegeneracy.
- Do not introduce a common codimension-one-submanifold superclass. Any smoothness/codimension theorem
carries its regularity/transversality hypotheses; equal-phase adapters carry nonzero/sign conditions.
- Keep existing wall/local-finiteness results with explicit comparison theorems after relabeling; actual
destabilization requires categorical existence data. No theorem is strengthened by terminology alone.
- id: mathematical-ownership-t4
gh_issue: 1315
kind: task
parent: mathematical-ownership-m1
title: MO1.04 — Separate hyperbolic-extension algebra from geometric Mukai realizations
status: done
summary: 'Extract reusable extension, reflection and Gram-form algebra from Lattice/Mukai while retaining
geometric Mukai terminology for the realization and its geometric consequences. Native prerequisites
#1223 and #1229 avoid moving the shared root while its canonical form and bridges are still being
established. CA3 remains the sole owner of those mathematical deliverables.'
priority: must
size: L
tags:
- ownership
- architecture
- review-03
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:algebra
- area:algebraic-geometry
- type:refactor
- sev:high
- ready
acceptance:
- Identify the existing Z × N × Z carrier and its real extension as hyperbolic-extension algebra. Place
general reflections and Gram theory at their neutral lattice/form owners with explicitly sufficient
hypotheses.
- Move or adapt IsSpherical, expectedDim, Mukai-vector/Euler interpretations and geometric period identifications
at the application boundary; preserve the distinctions between a numerical model and an integral surface
Mukai lattice.
- 'Consume #1223’s coefficient-ring generalization of the EXISTING Mukai.pairing, keeping fixed arity
three and its required public name. Do not create a second generalized pairing because the directory
changes.'
- 'Preserve #1229’s n=2, weighted rank-one and factor-of-two bridges. Check the actual carriers: HDeg
2 is a Fin-indexed function type and RealExtension is a product; supply the explicit coordinate equivalence
when needed rather than assuming definitional equality.'
- Preserve the odd-degree alternating obstruction, the polarization weight and the loss of information
at higher divisor rank. Integrality and Euler-pairing identifications are proved or supplied geometric
inputs.
- 'Treat #1230’s pairCharge and any graded pairing as separate decisions. #1230 was withdrawn on 2026-09-30; no speculative graded parent is needed for this extraction.'
blocked_by:
- 1312
- 1223
- 1229
- id: mathematical-ownership-t5
gh_issue: 1316
kind: task
parent: mathematical-ownership-m1
title: MO1.05 — Remove numerical parent-to-specialization import inversions
status: done
summary: 'Split generic numerical/Todd/transport foundations from K3, dimension-specific and cross-surface
comparison files. Structural continuation of closed #1221, coordinated with open #1222 and #1225.
Those issues retain the correction/fourfold constructions; this one removes backwards imports.'
priority: must
size: M
tags:
- ownership
- architecture
- review-07
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:algebraic-geometry
- area:stability
- type:refactor
- sev:high
- ready
acceptance:
- Split Numerical/Mukai/SqrtTodd so generic correction algebra does not import K3 Riemann–Roch. Put
K3 identification downstream, preserving the explicit codimension-at-most-four scope of sqrtComp.
- 'Split generic PolarisedWallTransport from surface/threefold comparisons: the generic (n,m,κ) root
imports no specialized transport, and each comparison imports both sides.'
- Separate Abelian/K3 and Enriques/Abelian numerical comparison material so each model imports the common
foundation rather than a sibling merely to reuse a comparison.
- 'Preserve #1221’s transport, #1222’s arbitrary-surface rank-one pointwise √td comparison and all parameter
reindexing obligations. n and m remain separate; κ=1 and κ=√td are distinct choices.'
- Add focused transitive-import checks for each extracted root and verify at least the existing surface
and threefold consumers use it. Do not replace missing geometry with an unrestricted instance.
blocked_by:
- 1313
- id: mathematical-ownership-t6
gh_issue: 1317
kind: task
parent: mathematical-ownership-m1
title: MO1.06 — Distinguish numerical models, named surface cases and geometric realizations
status: done
summary: 'Organize reusable Numerical/Examples material as mathematical models and put variety-specific
realizations beneath the common numerical construction. Follows MO1.05’s comparison split; coordinates
CA1 #1225. The classification includes naming, import owners and realization maps, not a new numerical
K-theory.'
priority: should
size: M
tags:
- ownership
- architecture
- review-12
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:algebraic-geometry
- type:refactor
- sev:medium
- ready
acceptance:
- Classify each current Examples file as a reusable model, a comparison or an actual worked example;
move the first two to subject owners, leaving Examples for demonstrations.
- Choose a coherent surface/dimension model hierarchy and make K3, Enriques and abelian surface models
siblings. Move the K3-only central-charge material out of the vaguely generic Numerical/GrothendieckGroup
owner.
- Keep actual schemes/surfaces and their realization maps distinct from formal rank-degree-coordinate
models; record the map that makes each advertised specialization reach the common root.
- Label the Enriques rank-one model as a polarization slice, not the full numerical lattice. Do not
infer higher-dimensional numerical-ring generation by H from Picard rank one.
- Preserve arbitrary-divisor-rank charges as a separate input branch of the shared exponential kernel,
not as a child of lossy scalar H-degree compression.
- 'Document where the reusable fourfold models of #1225 belong, without reproducing their charge formula
or claiming a geometric existence theorem.'
blocked_by:
- 1316
- id: mathematical-ownership-t7
gh_issue: 1318
kind: task
parent: mathematical-ownership-m1
title: MO1.07 — Move linear Serre duality and Yoneda helpers to their independent owners
status: done
summary: 'Make the basic Serre-functor and linear representability theory importable without triangulated
categories. SRF1 #897–#899 remains responsible for new full-faithfulness, transport and geometric-duality
theorems. This issue relocates/repackages the owner; it does not duplicate those proofs.'
priority: must
size: M
tags:
- ownership
- architecture
- review-05
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:derived
- type:refactor
- sev:high
acceptance:
- Move SerreFunctor/Basic and linear uniqueness/equivalence material to CategoryTheory/Linear/SerreFunctor
under the agreed map; retain shift/Ext/triangulated results in downstream adapters.
- Extract isoOfLinearYonedaIso and the independent hom_ext helper to a linear Yoneda owner; the Serre
API consumes it.
- Preserve public declaration names and assumptions; full faithfulness still needs Hom-finiteness and
essential surjectivity is not inferred from full faithfulness.
- 'In #898 separate ordinary linear conjugation transport from shift coherence, exactness and TriEquiv
packaging if they land together. No shift/exactness assumption is added to the linear core.'
- Coordinate the open SRF1 path contracts and audit routing with the move. Show a minimal linear-core
import has no path to Triangulated or stability.
blocked_by:
- 1312
- id: mathematical-ownership-t8
gh_issue: 1319
kind: task
parent: mathematical-ownership-m1
title: MO1.08 — Separate abelian stability foundations and make slope theory independent of Gieseker
status: done
summary: 'Extract abelian stability functions/HN foundations from the triangulated weak-stability application
and organize geometric slope and Gieseker theories as siblings. SF1 #192/#208 owns weak stability
and HRS tilting; CA2 #1226–#1228 owns the tilt-charge comparisons. Neither set of mathematical deliverables
is refiled here.'
priority: must
size: L
tags:
- ownership
- architecture
- review-06
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:stability
- area:algebraic-geometry
- type:refactor
- sev:high
- ready
acceptance:
- Move the independently abelian StabilityFunction/HN declarations from Weak/Foundation/StabilityFunction
to the agreed CategoryTheory/Abelian stability owner; triangulated weak stability consumes them.
- Retain Weak as a legitimate child of StabilityCondition. The task is the misplaced abelian root, not
reversing every weakened-concept directory.
- Separate μ-stability and μ-HN from the Gieseker owner, using sibling Slope and Gieseker modules with
downstream comparison files.
- Preserve weak versus ordinary stability, the zero-charge subcategory, chain/HN hypotheses and geometry
firewall; do not identify slope and Gieseker stability.
- Demonstrate independent abelian and slope imports and preserve current tilting and geometric consumers.
Coordinate audit routing with SF1 rather than introducing a parallel stability-function carrier.
blocked_by:
- 1312
- id: mathematical-ownership-t9
gh_issue: 1320
kind: task
parent: mathematical-ownership-m1
title: MO1.09 — Own H⁰ constructions with DGCategory and make enhancement exactness explicit
status: done
summary: 'Separate DGCategory’s intrinsic H⁰/shift/triangle/functor theory from the data comparing it
with a chosen triangulated category. DG3 #854/#855 owns seam and exactness proofs. MO1 owns module
placement and truthful comparison packaging; later DG4/DG5 consumers must see the distinction.'
priority: must
size: L
tags:
- ownership
- architecture
- review-15
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:dg
- area:derived
- type:refactor
- sev:high
- ready
acceptance:
- Move generic H⁰ shift, triangle and pretriangulated compatibility results from Triangulated/DGEnhancement/H0
to Algebra/Homology/DGCategory/H0 or Pretriangulated as the declaration owner requires.
- Keep the present DGCategory on HomComplex; no enriched-category reencoding is part of this issue.
Preserve identity/composition and the existing DGFunctor.h0 construction.
- Distinguish an ordinary equivalence H⁰(C) ≌ D from an exact enhancement of triangulated D. If retaining
the current weaker carrier, label it as an underlying presentation and add the exact comparison/refinement
explicitly.
- 'Use #855 for linear/exact H⁰ and #854 for the classical seam; do not replace their missing theorem
with a supplied Boolean or exactness field that is advertised as proved.'
- Keep object-specific homotopy-category realizations with HomotopyCategory. Intrinsic dg H⁰ imports
no enhancement consumer, scheme realization or stability.
- Preserve supplied-vs-proved boundaries and avoid any general uniqueness-of-enhancements assertion.
blocked_by:
- 1312
- id: mathematical-ownership-t10
gh_issue: 1321
kind: task
parent: mathematical-ownership-m1
title: MO1.10 — Extract derived tensor and pushforward capabilities from Fourier–Mukai
status: done
summary: 'Give general derived operations their own geometry-level owners, leaving Fourier–Mukai files
to assemble kernels, convolution and transforms. DT1 remains owner of tensor inhabitation; #795/#796
remain owners of transform equivalence/adjoint obligations; SF8 #554 owns arbitrary pullback. This
is a placement split, not discharge of those trust boundaries.'
priority: must
size: L
tags:
- ownership
- architecture
- review-10
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:derived
- area:algebraic-geometry
- type:refactor
- sev:high
- ready
acceptance:
- Extract HasDerivedTensor, HasDerivedPushforward and the general coherent derived-tensor compatibility
from KernelCorrespondence/DerivedTensorCoherence to agreed DerivedCategory/Tensor and pushforward
owners; separate unbounded, bounded-coherent and relative inputs.
- Retain kernel-specific projection formula, convolution and compositor/adjunction data with FourierMukai;
avoid mechanically moving every class mentioned by a transform.
- 'Review ordinary-category Correspondence separately: move it to a neutral KernelTransform owner only
if the actual declarations require no triangulated structure. No speculative common carrier.'
- 'Coordinate DT1 #892/#928–#931 constructors and path contracts with the extracted existing roots.
Do not recreate their bifunctor, resolution interface or coherence.'
- Preserve finite-Tor/boundedness and exactness hypotheses. Derived tensor on bounded coherent objects
is conditional; singular schemes do not acquire it through relocation.
- Demonstrate a general tensor/pushforward import without FourierMukai or stability, and rebuild the
transform/relative-family consumers against the same capabilities.
blocked_by:
- 1312
- id: mathematical-ownership-t11
gh_issue: 1322
kind: task
parent: mathematical-ownership-m1
title: MO1.11 — Move flatness and relative-perfect predicates out of the moduli consumer
status: done
summary: 'Make geometric flatness, pseudo-coherence and Tor-amplitude predicates usable independently
of the relative-perfect moduli problem. SF8 #517/#554 and #723 retain construction, preservation and
compact-perfect obligations. The issue is foundational ownership, with the existing trust boundary
unchanged.'
priority: must
size: M
tags:
- ownership
- architecture
- review-11
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:moduli
- area:derived
- type:refactor
- sev:high
acceptance:
- Extract module flatness over a scheme morphism from Moduli/PerfectComplex/Relative to Modules/Flat
(or the actual module owner agreed in MO1.01).
- Place pseudo-coherence, Tor amplitude and intrinsic/relative perfect-complex predicates with DerivedCategory/Perfect;
the moduli groupoid and descent consume those owners.
- 'Keep the current schemePseudoCoherent criterion explicitly restricted/interpreted as in #554: on
arbitrary non-Noetherian bases it is not automatically standard pseudo-coherence. A move must neither
erase this warning nor pretend to prove the missing preservation theorem.'
- Preserve the distinction and one-way comparisons among schemePerfect, schemeRelativePerfect and TwoTermPerfectDeterminantData;
add no global equivalence instance.
- Rebuild existing relative-perfect, determinant and moduli consumers, and show the foundational predicates
have no transitive Moduli/stability dependency.
blocked_by:
- 1312
- id: mathematical-ownership-t12
gh_issue: 1323
kind: task
parent: mathematical-ownership-m1
title: MO1.12 — Separate the GL⁺(2,ℝ) cover from its stability action
status: done
summary: 'Extract the standalone covering-group/topological construction currently owned by stability
symmetry. Related AUT1 #927 concerns an autoequivalence action and supplied proper discontinuity;
it is not the owner of this independent GL-cover construction and is not discharged here.'
priority: should
size: M
tags:
- ownership
- architecture
- review-08
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:stability
- type:refactor
- sev:medium
- ready
acceptance:
- Classify GLTilde declarations into group/cover theory and phase/slicing/action adapters. Place the
former by the pinned general-linear/covering API under the approved map.
- Retain stability-specific phase lifts and action compatibility downstream; do not infer a covering
theorem from the name GLTilde if it is only supplied in the current formalization.
- Preserve public names and the current action laws; prove or retain the explicit projection/comparison
consumed by stability.
- Show the cover-core import reaches neither StabilityCondition nor AlgebraicGeometry, and verify the
existing symmetry consumer still elaborates.
blocked_by:
- 1312
- id: mathematical-ownership-t13
gh_issue: 1324
kind: task
parent: mathematical-ownership-m1
title: MO1.13 — Make mass a sibling of metric and extract planar convex geometry
status: done
summary: 'Give mass its own stability-theoretic owner and move neutral planar perimeter/convexity lemmas
out of a mass-subadditivity proof directory. Relates to the mass-Hom consumer #1185, whose theorem
and moduli input remain outside this structural issue.'
priority: should
size: M
tags:
- ownership
- architecture
- review-09
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:stability
- type:refactor
- sev:medium
- ready
acceptance:
- Place mass definitions and intrinsic mass results under StabilityCondition/Mass, with Metric importing
the mass API it uses.
- Extract Euclidean polygon/perimeter/convex geometry from Metric/Mass/Subadditivity to the nearest
appropriate neutral analysis/convexity owner, checking actual Mathlib definitions first.
- Keep HN-polygon and semistability interpretations in stability adapters; show the planar core imports
no categories or stability.
- Retain mass triangle/subadditivity and metric consumers with the same hypotheses and normalizations;
do not reconstruct mass under a second name.
- Give proof-shaped or paper-number files in this touched subtree a descriptive subject owner while
retaining citation/source bindings.
blocked_by:
- 1312
- id: mathematical-ownership-t14
gh_issue: 1325
kind: task
parent: mathematical-ownership-m1
title: MO1.14 — Repair direct Mathlib-owner mismatches in module and sheaf APIs
status: done
summary: Perform the small, independently reviewable owner corrections where the current path disagrees
with the API being extended. These are direct owner corrections, not a wholesale reorganization of
the broadly sound sheaf/site/scheme hierarchy.
priority: should
size: M
tags:
- ownership
- architecture
- review-13
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:algebra
- area:algebraic-geometry
- type:refactor
- sev:medium
acceptance:
- Move CategoryTheory/ModuleCat/LinearDual to Algebra/Category/ModuleCat beside the existing exactness
API.
- Split/move the presheaf exterior-power content currently under ModuleCat/Sheaf to the actual PresheafOfModules
owner; retain genuinely sheaf-specific declarations with sheaves.
- Place the quasicoherent extension-closure theorem in Modules/Quasicoherent; isolate any affine/cohomological
helpers to avoid a cycle rather than retaining a proof-technique owner for the theorem.
- Place the scheme-module exterior-power restriction API currently under Divisors/ExteriorPower with
Modules/ExteriorPower/Restriction, leaving divisor-specific consequences downstream.
- Use separate small PRs if import closures differ. Preserve names, audits and source bindings, and
verify each destination against the pinned Mathlib API rather than a guessed upstream path.
blocked_by:
- 1312
- id: mathematical-ownership-t15
gh_issue: 1326
kind: task
parent: mathematical-ownership-m1
title: MO1.15 — Verify the completed ownership graph and close the migration ledger
status: done
progress: 'Recorded a disposition for all fifteen findings in
docs/architecture/cutover-ledger.md, each with its implementing pull request, its
agreed owners and the gate that now holds them. Every row is implemented; none was
closed by adjudication alone. The closure found one real gap: of the nine cores this
task names, eight carried a numbered layering rule and perfectness (row 11, MO1.11)
carried only a RETIRED_PATHS entry, which asserts the old path is gone and nothing
about what the new owners may import. Rule 18 pins both edges transitively -- the
owners reach neither the moduli nor the stability tree, and Moduli/PerfectComplex
still reaches them -- with allowed and forbidden known-answer fixtures. Rows 03 and
13 keep retired-path coverage deliberately; neither is among the nine, and the
acceptance asks for fixtures on genuine policy edges rather than a subject ranking.
No CA, SRF1, DG3, DT1 or SF8 obligation is discharged.'
summary: Close MO1 on evidence that the extracted foundations are reusable and every review finding
has a completed or explicitly adjudicated disposition. This is the integration acceptance task for
the companion milestone, not another theorem-development track.
priority: must
size: M
tags:
- ownership
- architecture
- review-14
labels:
- initiative:architecture-hardening
- arch:layering
- area:lean
- area:governance
- type:task
- sev:medium
- ready
acceptance:
- Record a disposition for findings 01–15 with the implementing PR, old/new owner and canonical specialization/realization
map. No finding is dropped because a path merely looks cleaner.
- Run focused transitive-import checks for neutral pairing/charge cores, generic numerical transport,
linear Serre/Yoneda, abelian stability, dg H⁰, derived operations, perfectness, GL cover and planar
geometry. Add regression fixtures for genuine policy edges, not a total subject ranking.
- Ensure umbrellas expose neutral cores without reimporting specializations; update coverage exceptions,
single-owner checks, audit routing, registry/source-owner paths, historical-name executable support
and CI paths together with each move.
- Verify CA1–CA3 retain their canonical roots, comparison obligations and independent state. New paths
must not reopen completed kernel work, duplicate pending discriminant/tilt proofs or convert the conditional
graded-root proposal into a mandatory structure.
- Run check_roadmap.py --require-api and all relevant architecture checks, then obtain the prescribed
full runner CI on the final implementation commit. No local full-library build, hidden axiom or sorry
is a refactoring shortcut.
- The milestone closes only after all accepted placement work has landed; any rejected proposal needs
a documented mathematical/import-graph reason and tracker/roadmap agreement. Completion does not assert
that unrelated SRF1, DG3, DT1, SF8 or all CA mathematics is finished.
blocked_by:
- 1314
- 1315
- 1317
- 1318
- 1319
- 1320
- 1321
- 1322
- 1323
- 1324
- 1325Exported from .claude/roadmap/mathematical-ownership.yaml at the commit that removed it from the tree.
Architecture
- Canonical roots and specialization trees
- Structural cutover ledger
- Generalization backlog
- Dependency direction
- Loop-engineering friction log
- Mathematical ownership and abstraction boundarie
- Declaration placement
- When a proposition-valued class is justified
- Enhanced spherical twists: conventions and remai
- Verification history
CI and runners
- Bounded audit batching probe, 2026-09-27
- CI1.10: audit and exception record contention, 2
- Proposed hosted-only CI cutover for CI1.07
- Mac VM runners
- Owner CI lane: self-hosted runners in a private
- CI performance baseline, 2026-09-27
- CI1.07 exact-tree Ubuntu routing sample, 2026-09
- Ubuntu workstation runners (historical topology)
Decision records
- ADR-0001 — The seam is cold: files, not runtime (superseded)
- ADR-0002 — Identity is minted in the topic repo, (superseded)
- ADR-0003 — Elaboration and axiom evidence is pro (superseded)
- ADR-0004 — The statement registry lives in the t (superseded)
- ADR-0005 — Seven axes, four values, no aggregate (superseded)
- ADR-0006 — All contract issues are filed on `bri (superseded)
- ADR-0007 — Where the contract package lives (superseded)
-
ADR-0008 —
@[cites]ships as a shared Lake dep (superseded) - ADR-0009 — The contract package lives with the e (superseded)
- ADR-0010 — How a dg category is encoded, and whe
- ADR-0011 — Hom-complexes are complexes of abelia
- ADR-0012 — One Mathlib-style source root
- ADR-0013 — Own reusable moduli infrastructure in
- ADR-0014 — Stacks, algebraic spaces and moduli:
Roadmaps