Versioned Protected-Data Exports and Cross-Volume Round Trips

Protected data can move safely between subjects only when the exported object carries its theory definition, protection mechanism, conventions, scope, provenance, uncertainty, semantic version, and content hash. A bare list of dimensions or central charges invites normalization drift and stronger downstream claims than the source supports. This page defines a compact, content-addressed record and a round trip that preserves those boundaries, following the general findability and provenance discipline of the FAIR principles Wilkinson et al. 2016.

Required background. Use the normalization discipline of the AGT dictionary and the information-loss analysis of the four-dimensional chiral-algebra sector.

Helpful background. Protected SCFT data for bootstrap use shows how a neighboring subject consumes only a selected subset.

The export is a typed scientific object

A protected-data record should answer five questions without consulting prose:

1. What theory? Dimension, supersymmetry, global form, deformation point, and any defect or background.
2. What quantity? Operator, algebra, anomaly, index coefficient, partition function, or category-level datum.
3. Why exact? Shortening, Q-cohomology, anomaly, localization, holomorphy, or a stated combination.
4. In which convention? Generator normalization, charge basis, scale, counterterms, fugacities, and parameter map.
5. How far does it reach? Included claims, excluded inferences, uncertainty, unresolved fields, and source cutoff.

The minimal field groups are:

Field group	Required content
theory	Human-readable identity and complete global/background qualifiers
observable	Kind, sector, arguments, units, and normalization
values	Exact expressions or numerical values with uncertainty
protection	Mechanism for each exported value
conventions	Charge, anomaly, operator, and scheme conventions
scope	Explicit inclusions and exclusions
sources	Stable routes or primary references with locators
source_cutoff	Latest evidence date represented by the record
unresolved	Named unknowns with reasons, never silent nulls
versions and hash	Schema compatibility and content identity

A canonical Argyres–Douglas example

The following record exports a bounded subset of the $(A_1,A_2)$ theory card. Rational numbers are strings so no binary floating-point conversion occurs.

{
  "schema_version": "1.0.0",
  "record_version": "1.0.0",
  "theory": {
    "dimension": 4,
    "name": "(A1,A2) Argyres-Douglas SCFT",
    "supersymmetry": "N=2"
  },
  "observable": {
    "kind": "protected_theory_card",
    "sectors": ["Coulomb", "Schur"]
  },
  "values": {
    "a": "43/120",
    "c": "11/30",
    "c_2d": "-22/5",
    "delta_u": "6/5"
  },
  "protection": ["N=2 shortening", "Weyl anomaly", "Q-cohomology"],
  "conventions": {
    "chiral_algebra": "c_2d = -12 c_4d",
    "weyl_anomalies": "a,c normalized so a free N=2 vector has (5/24,1/6)"
  },
  "scope": {
    "includes": [
      "rank-one Coulomb data",
      "conformal anomalies",
      "chiral-algebra central charge"
    ],
    "excludes": ["long-multiplet spectrum", "generic OPE coefficients"]
  },
  "sources": [
    "/supersymmetry-duality/n2-seiberg-witten/argyres-douglas-class-s-interfaces/",
    "/supersymmetry-duality/protected-operators-defects-correspondences/chiral-algebra-correspondence/"
  ],
  "source_cutoff": "2026-08-10",
  "uncertainty": {"kind": "exact"},
  "unresolved": [],
  "content_hash": {
    "algorithm": "sha256",
    "canonicalization": "RFC 8785",
    "excluded_field": "content_hash",
    "value": "9a46b62d218a75be4824188c2c2ba6e01f4fd6dae50841f7e0652866dc2e024e"
  }
}

The hash is computed from the RFC 8785 canonical JSON representation after removing the content_hash field. It binds values, conventions, scope, sources, and cutoff together. Reordering object keys or changing whitespace leaves the canonical hash fixed; changing an array order or scientific value changes it RFC 8785, §§3–4.

The record contains enough redundancy for scientific checks:

$$-12c=-12\left(\frac{11}{30}\right)=-\frac{22}{5}=c_{2d},$$

and the Coulomb dimension obeys the rank-one relation

$$2a-c=\frac{2\Delta_u-1}{4}=\frac7{20}.$$

Schema validity alone would not catch a wrong sign in either identity, so semantic checks accompany structural validation.

Semantic versions and compatibility

Use semantic versions according to the meaning of the record:

• increment the patch version for source-locator or explanatory corrections that do not change values or interpretation;
• increment the minor version for backward-compatible added fields or new protected values;
• increment the major version when a normalization, field meaning, theory specification, or required schema changes incompatibly.

Every scientific change produces a new content hash even when it is backward compatible. Version numbers describe compatibility; hashes identify exact content. Neither substitutes for the other.

If an exact value is corrected, retain the superseded record and mark the replacement relationship. Silent mutation breaks reproducibility for every downstream calculation that cited the old hash.

The cross-volume round trip

A consumer should import the tuple

$$(\text{schema version},\text{record version},\text{content hash})$$

and return a derived record containing:

• the source tuple unchanged;
• the transformation name, version, parameters, and code or analytic formula;
• the subset of source fields actually used;
• derived values with propagated uncertainty;
• any information discarded by the transformation;
• the strongest allowed downstream claim.

For example, a conformal-bootstrap calculation may use $c$, flavor levels, and selected protected OPE data. It must not claim that the source supplied long-operator dimensions if the source scope excludes them. When a newer source hash appears, the consumer can determine mechanically whether its result is stale and whether the changed fields were among those used.

Provenance should describe entities, activities, and derivations rather than only naming a file. This structure follows the same separation formalized by the W3C provenance model W3C PROV-DM 2013, §§2–3.

Validation before publication

Run four layers of checks:

1. Structural: required keys, types, enumerations, and version syntax.
2. Referential: every source route and external locator resolves to the intended object.
3. Semantic: dimensions, charge bases, anomaly identities, exact relations, and limiting cases agree.
4. Cryptographic: canonicalize independently and reproduce the declared digest.

Then simulate one consumer round trip. Change a source field deliberately and confirm that the hash changes and the consumer marks its derivative stale. Change only JSON whitespace and confirm that the canonical hash does not change.

Failure modes

Hidden normalization. A central charge without the free-field convention can differ by a fixed factor across subjects.

False exactness. A numerical fit or conjectural correspondence must carry uncertainty and status rather than being serialized as an exact rational.

Unresolved mixing erased. Export a protected combination or equivalence class when individual operators are not separated.

Scope strengthened downstream. A protected algebra does not export a full spectrum, and an index does not export unsigned multiplicities.

Hash refreshed without review. A new digest confirms new bytes, not scientific correctness. Semantic checks must pass first.

Stale sources hidden. The source cutoff and exact source locators make later corrections discoverable.

Exercises

Verify the two internal identities in the example record.

Solution

First, $-12(11/30)=-132/30=-22/5$. Second, $2(43/120)-11/30=86/120-44/120=42/120=7/20$, while $(2(6/5)-1)/4=(12/5-1)/4=7/20$.

References

• Moreau, L., and P. Missier, eds. PROV-DM: The PROV Data Model. W3C Recommendation, 30 April 2013. W3C.
• Rundgren, A., B. Jordan, and S. Erdtman. JSON Canonicalization Scheme (JCS). RFC 8785, June 2020. RFC Editor.
• Wilkinson, M. D., et al. “The FAIR Guiding Principles for Scientific Data Management and Stewardship.” Scientific Data 3 (2016): 160018. DOI.