Skip to content

Glossary

Terms you will encounter in Wardline output, documentation, and configuration.

Term Definition
Analysis level Depth of taint tracking: L1 (function-level), L2 (variable-level), L3 (call-graph). Higher levels find more violations but scan slower. See Analysis Levels.
Annotation group A numbered grouping (1-17) of related wardline decorators that enforce a cohesive set of contracts. Groups 1-4 are framework-mandated (authority tier flow); groups 5-17 are supplementary. See Supplementary Groups.
Assured Taint state for Tier 2 data — structurally and semantically validated. See Taint States.
Authority tier One of four trust levels (1=highest to 4=lowest) assigned to code and data. See Taint States.
Boundary A function decorated with @validates_shape, @validates_semantic, @validates_external, or @restoration_boundary that promotes data from a lower tier to a higher tier.
Call edge A directed link from one function to another in a call graph, representing a function call. Used by L3 analysis to propagate taint interprocedurally.
Coherence check Cross-reference validation run by wardline coherence — verifies manifest, decorators, exception registry, and fingerprint baseline are mutually consistent.
Conformance gap A known deviation from Wardline specification requirements. Tracked in SARIF output (wardline.conformanceGaps) to signal where scanner behaviour does not fully conform to normative rules.
Control law Enforcement state of the scanner: normal (full enforcement), alternate (degraded but running), or direct (manifest unavailable, minimal enforcement). See Degradation condition for what triggers alternate/direct.
Convergence bound A safety limit on the number of fixed-point propagation iterations in L3 analysis. Exceeding it emits the L3-CONVERGENCE-BOUND diagnostic.
Coverage ratio Fraction of scanned functions with a statically-known taint state (0.0-1.0). Values below 0.8 indicate manifest gaps. Reported in SARIF as wardline.coverageRatio.
Decorator A Python @decorator from wardline.decorators that annotates a function with trust-boundary metadata. See Decorator Vocabulary.
Degradation condition A reason the control law becomes alternate or direct. One of: conformance_gaps_present, manifest_unavailable, ratification_overdue, rules_disabled, stale_exceptions_present.
Exception A recorded decision to accept a finding that would otherwise block the scan. Managed via wardline exception. See Governance.
Exceptionability Whether a finding can be excepted: UNCONDITIONAL (no), STANDARD (with approval), RELAXED (easily), TRANSPARENT (auto-suppressed). See Severity Matrix.
Expedited ratio Fraction of active exceptions granted via the expedited governance path (0.0-1.0). Capped at 15% by coherence checks; exceeding it signals governance bypass.
External_raw Taint state for Tier 4 data — untrusted external input. See Taint States.
Finding A single violation or diagnostic emitted by the scanner. Contains rule ID, severity, location, taint state, and exceptionability.
Fingerprint AST-based hash of a function's structure. Used to detect when code changes under an existing exception.
Fixed-point propagation The iterative algorithm that propagates taint through function call chains in L3 analysis until no taint assignments change. Subject to the convergence bound.
Gate-blocking A finding with ERROR severity and no active exception. Gate-blocking findings cause scan exit code 1. The count is wardline.gateBlockingCount in SARIF output.
Governance path The approval route for an exception: standard (requires reviewer approval) or expedited (fast-track, subject to the 15% expedited ratio threshold).
Governance profile Project-level policy setting: lite (fewer mandatory fields) or assurance (all governance fields mandatory). See Profiles.
Guarded Taint state for Tier 3 data — shape-validated but not semantically verified. See Taint States.
Integral Taint state for Tier 1 data — audit-critical, highest trust. See Taint States.
Join The operation that combines two taint states when data flows merge. Produces the least-trusted common state. See Taint States.
Known validator A function (declared in the manifest or detected heuristically) that unconditionally raises on invalid input. Used by PY-WL-008 to recognise two-hop rejection paths.
Manifest The wardline.yaml file that declares which modules belong to which tier, governance profile, and rule overrides. See Manifest Reference.
Mixed_raw Taint state produced when incompatible taint states merge — the absorbing element of the join lattice. See Taint States.
Overlay A per-directory wardline.overlay.yaml file that extends the root manifest with local boundary declarations and rule overrides. See Manifest Reference.
Qualname Fully-qualified Python name of a callable (e.g., myapp.handlers.auth.handle_login). Used in findings, exceptions, and wardline explain.
Ratification Formal sign-off of manifest metadata by an authorised person. Includes ratification_date and review_interval_days. Overdue ratification degrades the control law.
Rejection path A branch in a validation boundary function that raises an exception or returns early on invalid input. Required by PY-WL-008.
Restoration boundary A @restoration_boundary decorator that re-promotes data to a higher tier with explicit evidence (structural, semantic, integrity, institutional).
Retrospective scan A scan run with --retrospective COMMIT_RANGE that analyses findings introduced during a degraded-law window. Results are marked with wardline.retroactiveScan: true in SARIF.
SARIF Static Analysis Results Interchange Format (v2.1.0). Wardline's output format for CI/CD integration. See SARIF Format.
SCC Strongly Connected Component — a maximal set of functions that call each other (directly or indirectly) in a call graph. L3 analysis computes SCCs to propagate taint efficiently.
Severity How a finding affects the scan exit code: ERROR (blocks, exit 1), WARNING (informational), SUPPRESS (hidden unless verbose). See Severity Matrix.
Severity cell A single entry in the 72-cell matrix: the (severity, exceptionability) pair for a given (rule, taint state) combination. See Severity Matrix.
Taint state The validation confidence level carried by data: one of 8 canonical states. See Taint States.
Temporal separation A governance requirement that the author and approver of enforcement artefact changes are different people, or approval is deferred to a retrospective review window. Required in assurance profile.
Tier Shorthand for authority tier (1-4). See Authority tier.
Transition A declared taint-state change on a boundary decorator, e.g., (EXTERNAL_RAW, GUARDED) on @validates_shape.
Two-hop rejection delegation PY-WL-008's ability to follow rejection paths through a second-hop function call (e.g., a boundary that delegates validation to a library function).
Unknown_raw / Unknown_guarded / Unknown_assured Taint states assigned when the scanner lacks sufficient evidence to determine full validation status. See Taint States.
Validation boundary See Boundary.
Verification mode A scan mode (--verification-mode) that produces deterministic output with no timestamps. Suitable for reproducible CI and diffable SARIF.

Further Reading