notes

Log | Files | Refs

crdt.txt (922B)

      1 CRDT
      2 
      3 https://www.bartoszsypytkowski.com/the-state-of-a-state-based-crdts/
      4 
      5 We can discrimitate CRDTs using two core categories: state-based (convergent) and operation-based (commutative) data types. No matter which one we talk about, they all consists of two parts: replication protocol and state application algorithms.
      6 
      7 - Replication Protocol
      8 - State Application Algorithms
      9 
     10 Merge operation must conform to three properties:
     11 
     12 1. Commutativity (x . y = y . x)
     13 
     14 2. and Asociativity ((x . y) . z = x . (y . z)) which means that
     15    we can perform out of order merge operations and still end up
     16    with correct state
     17 
     18 3. Idempotency (x . x = x), so we don't need to care about potential
     19    duplicates send from replication layer
     20 
     21 Those properties are not easy to guarantee, but you're going to see how far we can go only by using two basic operations, which meet those criteria:
     22 
     23 - union of two sets
     24 - maximum of two values