MDBX Roadmap and Acceptance Specification

Version: MDBX-1-DRAFT

Defines delivery order, completion criteria, and task decomposition for constrained models.

1. Delivery strategy

Build in layers, not UI-first:

1. data model and file format
2. storage engine and encrypted persistence
3. history and conflict detection
4. import/export bridge
5. benchmarks and recovery tests
6. reference UI and client integration

2. MVP scope

MVP MUST include:

• single .mdbx vault file
• SQLite baseline
• encrypted project/entry records
• attachment schema + at least one working storage mode
• Tiga modes
• append-friendly history
• basic conflict metadata
• snapshot/recovery checkpoint
• KDBX import
• KDBX export or clear bridge plan
• benchmark harness against KDBX
• a reference client implementation

3. Workstream grouping

Use tracks: format, storage, sync, security, migration, benchmarks, reference-client, docs.

4. Benchmark evidence

Publish at least:

• small-save latency
• vault open latency
• search latency
• attachment add latency
• sync delta after metadata edit
• sync delta after attachment rename
• sync delta after attachment content replacement

5. Test matrices

Recovery matrix should cover interrupted saves/sync, metadata corruption, chunk corruption, stale heads, concurrent edits, delete/recover flows. Compatibility matrix should cover old/new readers, unknown field retention, and KDBX roundtrip basics.

6. Rejection rules

Unacceptable plans include postponing project/attachment core models, skipping benchmarks, skipping recovery tests, or omitting KDBX migration path.