ADR Index

Architecture decision records document the choices that shape Ferrocat's public behavior. Use them when you need the reasoning behind catalog modes, runtime keys, PO compatibility boundaries, or durable API semantics.

Current records:

• 0001: Rust-Native, Performance-First Core
• 0002: Provide Both Owned and Borrowed Parse APIs
• 0003: Use Byte-Oriented Scanning and Separate Structure From Semantics
• 0004: Model msgstr by Semantics Instead of Always Using Vec<String>
• 0005: Treat ICU as the Canonical Model and Gettext as a Compatibility Bridge
• 0006: Separate the PO Core from the High-Level Catalog API
• 0007: Do Not Support previous_msgid History
• ADR 0008: Normalize Headerless PO Files on Write
• 0009: Use Versioned, Truncated SHA-256 Keys for Runtime Catalog Compilation
• 0010: Add a Locale-Resolved Compiled Catalog Artifact API
• 0011: Add an NDJSON Storage Format for High-Level Catalog Workflows
• 0012: Make the High-Level Catalog API ICU-Native by Default and Move Gettext Plurals Behind an Explicit Compat Mode
• 0013: Add a Rust-Native Catalog Combine API
• 0014: Keep Ferrocat Host-Neutral and Let Palamedes Own JS/TS Application Integration
• 0015: Add ICU Authoring Diagnostics Without Owning Runtime Formatting
• 0016: Add Progressive Semantic Message Metadata Around msgid
• 0017: Add Catalog Audit Reports Without Reintroducing Fuzzy Matching
• 0018: Add Machine-Translation Metadata to Catalog Entries