1. Executive Summary & Scale Metrics

Before Google Docs (originally acquired as Writely in 2006), document collaboration was a painful manual chore: emailing files back and forth (Quarterly_Report_v2_FINAL_bob_edits.docx) or using centralized check-in/check-out locks in Microsoft SharePoint that blocked all other team members while one person made edits.

Google Docs proved that dozens of people spread across different continents could type in the exact same paragraph at the exact same millisecond without document locks, race conditions, or corrupted text.

2. Requirements & Production Constraints

Functional Requirements#

1. Real-Time Collaborative Editing: Multiple users can simultaneously insert, delete, style, and annotate text in a single shared document.
2. Instant Local Typing (Zero Input Lag): When a user presses a key on their keyboard, the character must render on screen in 0 milliseconds. The UI cannot wait for a network round-trip to confirm a keystroke.
3. Presence & Cursor Tracking: Users see real-time colored collaborator carets and text selections moving across the screen.
4. Offline Capability & Replay: Edits made while disconnected on an airplane must merge cleanly upon reconnecting.

Non-Functional Requirements & The String Shifting Dilemma#

• The Divergence Problem: If Alice and Bob both have the string "CAT" and simultaneously make edits, network latency means Alice's edit will execute before Bob's on Alice's laptop, but Bob's edit will execute before Alice's on Bob's laptop.
• Mathematical Convergence Requirement: Regardless of the order in which network packets travel across the internet, all collaborators must eventually arrive at the exact identical character-for-character document state.

3. The Naive Design & Why It Collapses

[Alice: Doc = "HELLO"] ── Keystroke '!' ──> POST /doc (Full Body: "HELLO!")
[Bob:   Doc = "HELLO"] ── Keystroke '?' ──> POST /doc (Full Body: "HELLO?")
                                                   │
                                                   ▼
                                         [Central Web Server]
                                          (Last Write Wins)

4. Deep Architecture: Layer-by-Layer Walkthrough

Google Docs solves this using Operational Transformation (OT). Edits are treated not as static states, but as mathematical operations (Insert, Delete, Retain) transformed dynamically against concurrent operations.