Operational Transform Requires Server-Side Operation Sequencing

Two clients concurrently edit the same document position. Without coordination, applying both edits produces corrupted text because each operation was designed against a different document state.

OT uses a transform function that adjusts the position/content of a concurrent operation so it can be applied after another operation has already been applied. A central server serializes all operations and assigns revision numbers.

interface InsertOp {
  type: 'insert';
  position: number;
  text: string;
  revision: number;
  clientId: string;
}

// Server: transforms incoming op against all ops since client's revision
function transformAgainst(incoming: InsertOp, applied: InsertOp): InsertOp {
  if (applied.position <= incoming.position) {
    return { ...incoming, position: incoming.position + applied.text.length };
  }
  return incoming;
}

// Server applies and broadcasts the transformed op
function receiveOp(incoming: InsertOp, history: InsertOp[]): InsertOp {
  const concurrent = history.filter(op => op.revision >= incoming.revision);
  return concurrent.reduce(transformAgainst, incoming);
}

OT was designed for centralized architectures like Google Docs. The server is the single source of truth for operation order, which makes the transform function simpler but creates a hard dependency on server availability.