Skip to main content

Conformance

This section is normative.

The conformance framework defines what it means for tools, runtimes, agents, datasets, loggers, and analysis pipelines to be considered compliant with the MIND Standard. Conformance is profile-based to support heterogeneous workflows across XR, robotics, biosensing, CV, mocap, and embodied AI.

Profile-Based Conformance​

MIND uses a profile-based conformance model.

A conformant implementation MUST declare support for one or more Conformance Profiles.
Each profile defines:

  • required modalities
  • required metadata types
  • required serialization support
  • required features (timestamps, references, events)
  • required validation rules

An implementation MUST satisfy all requirements of each profile it claims.

Core Conformance Profile​

The Core Profile is the minimal requirement for any MIND dataset or tool.

A Core Profile implementation MUST:

  • Support JSON read/write
  • Support FlatBuffers read
  • Support all normative core modalities (Section 5)
  • Support the canonical timestamp model (Section 4)
  • Support required metadata (Section 7)
  • Support EventPoint + IntervalEvent (Section 6)
  • Enforce strict reference resolution
  • Preserve unknown optional extensions
  • Validate identifiers and schemas

Any dataset failing these rules is NOT a MIND-compliant dataset.

Extension Discovery and Capability Declaration​

Implementations MUST declare:

supported_extensions: [...]

Containers MAY declare:

required_extensions: [...]
optional_extensions: [...]

Rules:

  • A required extension that is unsupported MUST cause validation failure.
  • An unknown optional extension MUST NOT cause failure.
  • Unknown required extensions MUST cause failure.

Required vs Optional Features​

Containers MAY list:

required_features: [...]
optional_features: [...]

If a required feature is unsupported, the container MUST be invalid.

Feature identifiers MUST reference:

  • modalities
  • metadata
  • event types
  • container capabilities

Error Classes​

Validators MUST classify errors using the following codes:

  • INVALID_CONTAINER
  • SCHEMA_MISMATCH
  • UNRESOLVED_REFERENCE
  • TIMESTAMP_VIOLATION
  • EXTENSION_REQUIRED_UNSUPPORTED
  • METADATA_INVALID
  • INDEX_INVALID

Validators MAY define additional subclasses but MUST NOT override these canonical names.

Machine-Readable Conformance Tests​

MIND MAY include a test suite:

tests/
  valid/
  invalid/
  warnings/

Each test MUST:

  • provide a container
  • optionally provide external resources
  • specify the expected result (valid/invalid/warning)

Implementations SHOULD support executing these tests.

Versioned Conformance Profiles​

Profiles MUST be versioned:

  • MIND_CoreProfile@1.0.0
  • MIND_XRLoggingProfile@1.0.0
  • MIND_RealtimeAgentProfile@1.0.0
  • MIND_OfflineAnalysisProfile@1.0.0

Profile versions allow independent evolution.

Compatibility Levels​

Implementations MUST declare compatibility level:

Level 1 — Core​

Implements Core Profile.

Level 2 — Full​

Implements all normative features of Sections 3–8.

Level 3 — Complete​

Implements all extension-handling rules + advanced features (indexing, streaming, compression).

Compatibility level MUST be reported via:

implementation.compatibility_level

Forward and Backward Compatibility​

Rules:

Forward compatibility:​

  • Unknown fields MUST be preserved and ignored unless required.
  • Unknown optional extensions MUST NOT fail.
  • Unknown major version: MAY fail.
  • Unknown minor/patch version: MUST be accepted.

Backward compatibility:​

  • Deprecated fields MUST remain valid for one major version.
  • Newer containers SHOULD load on older tools if extensions are optional.

Summary​

Conformance ensures:

  • predictable validation
  • safe extension handling
  • version and schema stability
  • forward/backward compatibility
  • explicit profiles and feature negotiation
  • robust interoperability across XR, robotics, CV, biosensing, and agents