Permanence Architecture

CUID2-based immutability and namespace versioning designed to reduce migration burdens and enable long-term data persistence.

Dual-Layer Permanence Architecture

+-----------------------------------------------------+
|           Layer 2: Namespace Versioning              |
|  (Cross-Version Coexistence: SDC4 - SDC5 - SDC6)    |
+-----------------------------------------------------+
|  https://semanticdatacharter.com/ns/sdc4/            |
|  https://semanticdatacharter.com/ns/sdc5/            |
|  https://semanticdatacharter.com/ns/sdc6/            |
+-----------------------------------------------------+
         ^                 ^                 ^
         |                 |                 |
+--------+-----+  +-------+------+  +-------+------+
| Layer 1:     |  | Layer 1:     |  | Layer 1:     |
| CUID2-Based   |  | CUID2-Based   |  | CUID2-Based   |
| Immutability |  | Immutability |  | Immutability |
| (Within V4)  |  | (Within V5)  |  | (Within V6)  |
+--------------+  +--------------+  +--------------+

Layer 1: Immutable Structural Components

Problem Solved: Within-version model immutability - create new models without migrating old data

The Mechanism

SDC4 decouples structure from semantics using CUID2 identifiers:

  • mc-<cuid>: Model Component type definition (structure)
  • ms-<cuid>: Model Structure element (instance)
  • label: Fixed semantic meaning (e.g., "Systolic Blood Pressure")

Schema Example: Immutable Blood Pressure Component

<!-- Defined once in 2025, designed for long-term persistence -->
<xsd:complexType name="mc-clj5x1g8f000008l09j7f6c3d">
  <xsd:complexContent>
    <xsd:restriction base="sdc4:ClusterType">
      <xsd:sequence>
        <!-- Semantic meaning is FIXED -->
        <xsd:element name="label" type="xsd:string"
                     fixed="Blood Pressure" />

        <!-- Structure references immutable systolic component -->
        <xsd:element ref="sdc4:ms-clj5x2p4k000108l01a2b3c4d"
                     minOccurs="1" maxOccurs="1" />

        <!-- Structure references immutable diastolic component -->
        <xsd:element ref="sdc4:ms-clj5x2p4k000108l01a2b3c4e"
                     minOccurs="1" maxOccurs="1" />
      </xsd:sequence>
    </xsd:restriction>
  </xsd:complexContent>
</xsd:complexType>

<!-- Element declaration for substitution -->
<xsd:element name="ms-clj5x1g8f000008l09j7f6c3d"
             type="mc-clj5x1g8f000008l09j7f6c3d"
             substitutionGroup="sdc4:Item" />

Why This Prevents Migration

  1. Permanent Structure: mc-clj5x1g8f000008l09j7f6c3d always means the same physical structure
  2. Fixed Semantics: label="Blood Pressure" is immutable in the schema
  3. Queryable by CUID2: Applications query ms-clj5x1g8f000008l09j7f6c3d, guaranteed to find the same structure
  4. No Name Dependency: Unlike traditional XML where <bloodPressure> to <bp> breaks compatibility, CUID2 remains stable

Instance Data (2025)

<sdc4:dm-h7k2x... xmlns:sdc4="https://semanticdatacharter.com/ns/sdc4/">
  <sdc4:dm-label>Patient Vitals</sdc4:dm-label>
  <sdc4:dm-language>en-US</sdc4:dm-language>
  <sdc4:dm-encoding>UTF-8</sdc4:dm-encoding>

  <sdc4:ms-clj5x1g8f000008l09j7f6c3d>
    <sdc4:label>Blood Pressure</sdc4:label>
    <sdc4:ms-clj5x2p4k000108l01a2b3c4d>
      <sdc4:label>Systolic</sdc4:label>
      <sdc4:xdquantity-value>120</sdc4:xdquantity-value>
      <sdc4:xdquantity-units>
        <sdc4:label>Units</sdc4:label>
        <sdc4:xdstring-value>mmHg</sdc4:xdstring-value>
      </sdc4:xdquantity-units>
    </sdc4:ms-clj5x2p4k000108l01a2b3c4d>
  </sdc4:ms-clj5x1g8f000008l09j7f6c3d>
</sdc4:dm-h7k2x...>

Same Data, Still Valid Against SDC4 Schemas in 2035

An application in 2035 can query 2025 SDC4 data using the SDC4 namespace:

# Query by CUID2 - guaranteed stable structure
bp_elements = doc.findall(".//sdc4:ms-clj5x1g8f000008l09j7f6c3d",
                           namespaces={'sdc4': 'https://semanticdatacharter.com/ns/sdc4/'})

for bp in bp_elements:
    # Structure is immutable, so this always works
    systolic = bp.find(".//sdc4:ms-clj5x2p4k000108l01a2b3c4d//sdc4:xdquantity-value")
    print(f"Systolic: {systolic.text} mmHg")

Result: Data from 2025 is designed to work in 2035 with minimal migration burden

Layer 2: Namespace Versioning

Problem Solved: Cross-version coexistence (SDC4, SDC5, SDC6...)

The Mechanism

Each SDC major version gets its own XML namespace URI:

  • SDC4: https://semanticdatacharter.com/ns/sdc4/
  • SDC5: https://semanticdatacharter.com/ns/sdc5/ (when released)
  • SDC6: https://semanticdatacharter.com/ns/sdc6/ (future)

Coexistence in XML Schema

<!-- SDC4 Schema (permanent) -->
<xs:schema
  xmlns:sdc4="https://semanticdatacharter.com/ns/sdc4/"
  targetNamespace="https://semanticdatacharter.com/ns/sdc4/"
  xmlns:xs="http://www.w3.org/2001/XMLSchema">

  <xs:complexType name="XdQuantityType">
    <!-- SDC4 definition -->
  </xs:complexType>
</xs:schema>

<!-- SDC5 Schema (coexists with SDC4) -->
<xs:schema
  xmlns:sdc5="https://semanticdatacharter.com/ns/sdc5/"
  targetNamespace="https://semanticdatacharter.com/ns/sdc5/"
  xmlns:xs="http://www.w3.org/2001/XMLSchema">

  <xs:complexType name="XdQuantityType">
    <!-- SDC5 definition (enhanced) -->
  </xs:complexType>
</xs:schema>

<!-- Application schema can import BOTH -->
<xs:schema
  xmlns:sdc4="https://semanticdatacharter.com/ns/sdc4/"
  xmlns:sdc5="https://semanticdatacharter.com/ns/sdc5/"
  xmlns:xs="http://www.w3.org/2001/XMLSchema">

  <xs:import namespace="https://semanticdatacharter.com/ns/sdc4/"
             schemaLocation="sdc4.xsd"/>
  <xs:import namespace="https://semanticdatacharter.com/ns/sdc5/"
             schemaLocation="sdc5.xsd"/>

  <!-- Use SDC4 components for legacy data -->
  <xs:element name="LegacyPatient" type="sdc4:DMType"/>

  <!-- Use SDC5 components for new data -->
  <xs:element name="EnhancedPatient" type="sdc5:DMType"/>
</xs:schema>

RDF Graph Partitioning

# Query SDC4 data only
SELECT ?patient ?age
FROM NAMED <https://semanticdatacharter.com/ns/sdc4/>
WHERE {
  GRAPH <https://semanticdatacharter.com/ns/sdc4/> {
    ?patient sdc4:hasAge ?age .
  }
}

# Federated query across SDC4 AND SDC5
SELECT ?patient ?ageSource ?age
WHERE {
  {
    GRAPH <https://semanticdatacharter.com/ns/sdc4/> {
      ?patient sdc4:hasAge ?age .
      BIND("SDC4" AS ?ageSource)
    }
  } UNION {
    GRAPH <https://semanticdatacharter.com/ns/sdc5/> {
      ?patient sdc5:hasAge ?age .
      BIND("SDC5" AS ?ageSource)
    }
  }
}

# Aggregate across all versions
SELECT ?ageSource (COUNT(?patient) AS ?count)
WHERE {
  {
    GRAPH <https://semanticdatacharter.com/ns/sdc4/> {
      ?patient sdc4:hasAge ?age .
      BIND("SDC4" AS ?ageSource)
    }
  } UNION {
    GRAPH <https://semanticdatacharter.com/ns/sdc5/> {
      ?patient sdc5:hasAge ?age .
      BIND("SDC5" AS ?ageSource)
    }
  } UNION {
    GRAPH <https://semanticdatacharter.com/ns/sdc6/> {
      ?patient sdc6:hasAge ?age .
      BIND("SDC6" AS ?ageSource)
    }
  }
}
GROUP BY ?ageSource

Why This Prevents Cross-Version Migration

  1. Separate Namespaces: SDC4 and SDC5 have distinct URIs - no name collision
  2. XML Schema Imports: Applications can import both schemas simultaneously
  3. RDF Named Graphs: Triplestores partition by namespace, enabling federated queries
  4. W3C Standards Foundation: XML namespaces are a core W3C standard providing long-term stability

Result: SDC4, SDC5, SDC6... designed to coexist with minimal cross-version migration

Example: Healthcare Network 2025-2045

2025
Deploy SDC4

Patient demographics, vitals, lab results

Namespace: .../sdc4/

2028
SDC5 Releases

(enhanced genomic data types)

  • SDC4 data continues operating - no migration
  • Build new genomic models in SDC5
  • Reuse SDC4 patient demographics via copy/edit

Namespaces: .../sdc4/ + .../sdc5/

2030
Running in Parallel

20% SDC5 data (genomics), 80% SDC4 data (legacy vitals)

Federated SPARQL queries across both

2033
SDC6 Releases

(AI-enhanced semantic links)

  • SDC4 + SDC5 data still operating
  • Gradual SDC6 adoption for new use cases

Namespaces: .../sdc4/ + .../sdc5/ + .../sdc6/

2045
20 Years Later

Original 2025 SDC4 data: validates against SDC4 schemas, queryable via SPARQL

SDC4 models designed for long-term stability - minimal migration burden

Migration approach
Reduced burden
Benefit
Lower costs

Why Traditional Approaches Fail

Challenge Traditional XML/JSON SDC Permanence Architecture
Element Rename <bloodPressure> to <bp>
Breaks all XPath queries
Requires data migration
ms-clj5x1g8f... (immutable CUID2)
Query by CUID2, not name
No migration needed
Need New Structure Change schema - old data invalid
ETL pipeline to migrate
Create new model (new CUID2)
Old data + old model remain valid
Major Version Upgrade Forced migration project
Significant cost
v1 + v2 coexist
Reduced migration burden
20-Year Data Lifecycle 4-6 migrations
High total cost
Data loss risk each time
Minimal migrations
Reduced migration costs
Lower data loss risk

When the Metadata Moves

A dataset that carries a code and a link to the code's meaning depends on someone else keeping the link alive. Publishers reorganize their sites, retire files and put records behind logins, and they owe the people downstream no notice. Three US federal sources did exactly that while the FAIR Data Demo was being built from them:

  • NHANES. CDC moved the 2017-2018 survey documentation. The old page for each data file now returns a "Page Not Found" page with HTTP status 200, so a link checker reports it as working.
  • BRFSS. The 2022 codebook PDF that defines the survey's coded answers returns 404. CDC now publishes it only inside a zip file.
  • The NIH CDE Repository. The public record of a common data element no longer shows the LOINC code of each answer without a login.

Nothing about the data changed. What changed was the metadata a reader needs to understand it, and it changed in places the data's users do not control.

A Semantic Data Charter model binds that meaning into the published model itself: each component's label and description, every permissible value with its label, the units, ranges and data types, and a source's own conventions where it has them. Links to the publisher's pages stay in the model as provenance. A published model never changes, so a record made on it still validates against it and still says what it said when a link behind it stops answering.

Research data repositories call the goal resilience: the ability to make data accessible, recoverable, and useful under both normal operations and crisis conditions, and they warn that when it fails, most people will not notice until they go looking for something that is no longer there (Sutton-Kennedy, Crall, Trice and Gum, Repository Resilience in Practice, ESIP, 2026). Data is useful only while its meaning can be read. Repository resilience keeps the systems that serve the data running; binding the metadata to the data keeps its meaning readable when those systems, or someone else's, change.

Sources checked 26 October 2026.