CRDT
Conflict-free Replicated Data Types — mathematical data structures that can be replicated across multiple sites and merged without coordination, guaranteeing eventual consistency.
Summary
Conflict-free Replicated Data Types — mathematical data structures that can be replicated across multiple sites and merged without coordination, guaranteeing eventual consistency.
CRDTs are the theoretical foundation for multi-site object storage systems that need conflict-free convergence. When two sites independently modify metadata or object state, CRDTs ensure that merging produces a deterministic, consistent result without requiring distributed locks.
- CRDTs solve data structure convergence, not application-level conflicts. Two sites independently writing different content to the same S3 key is an application conflict that CRDTs alone cannot resolve.
- CRDT-based systems trade strong consistency for availability and partition tolerance. Not appropriate for workloads that require linearizable reads.
enablesActive-Active Multi-Site Object Replication — the mechanism for conflict-free multi-site convergencesolvesGeo-Replication Conflict / Divergence — mathematical guarantee of convergencescoped_toObject Storage — applicable to distributed object storage metadata
Definition
Conflict-free Replicated Data Types — a family of data structures mathematically guaranteed to converge to the same state when replicated across multiple sites, even with concurrent updates and no coordination.
Active-active geo-replicated object stores must handle concurrent writes to the same key at different sites. CRDTs provide a formal framework for designing metadata and state that merges deterministically without conflicts or consensus protocols.
Active-active object metadata replication, conflict-free bucket metadata convergence, distributed object store consensus avoidance.
Recent developments
- Yjs is the most-deployed CRDT library in the world (2026). Powers Figma multiplayer cursors, Linear's offline-first issue sync, Apple Notes cross-device sync, Obsidian Sync, and the Automerge ecosystem. YATA algorithm — significantly more memory-efficient than earlier CRDT designs. Per Taskade — History of CRDTs 2026: Lamport to Yjs, Automerge, Peritext.
- Yjs ecosystem: y-websocket, y-indexeddb, ProseMirror, Quill, CodeMirror, Monaco bindings. A full network + persistence + editor-binding stack — adopters wire CRDTs into existing rich-text editors without rewriting them. Per GitHub — yjs/yjs.
- Automerge remains the JSON-general-purpose alternative. Designed for arbitrary JSON-shaped data, not just text editing — broader applicability but less ecosystem mass than Yjs. Martin Kleppmann's 2017 work remains the canonical "CRDT for application developers" reference. Per arXiv 1608.03960 — A Conflict-Free Replicated JSON Datatype.
- Peritext (rich-text CRDT) named alongside Yjs/Automerge as 2026's three pillars. The CRDT field consolidated around three architectures: Yjs (text-first, web-deployed), Automerge (JSON-general, application-developer-ergonomic), Peritext (rich-text with stable formatting under concurrent edits). Per Taskade — History of CRDTs 2026.
- CRDTs vs Operational Transforms — CRDTs won. 2026 industry framing: OT was the 2000s answer; CRDTs are the 2020s answer; the production-collab-app cohort has fully migrated. OT survives mainly in Google Docs' legacy core. Per Fordel Studios — Real-Time Data Sync: CRDTs, OT, and What Actually Works.
- CRDT Dictionary published as field reference (November 2025). Ian Duncan's "CRDT Dictionary" catalogs the data-structure zoo — G-Set, OR-Set, LWW-Register, MV-Register, Sequence-CRDT, Counter-CRDT — as a working engineer's lookup table. Signals the field is mature enough for field-reference-style documentation. Per Ian Duncan — The CRDT Dictionary.
Connections 3
Outbound 3
scoped_to1Resources 2
CRDT community resource hub with links to papers, implementations, and educational materials on conflict-free replicated data types.
Shapiro et al. foundational paper defining CRDTs and proving their convergence properties, the theoretical basis for eventually consistent replication.