The Consensus Problem
Asha has exactly 5 coins. At the same moment she signs two transactions: one paying all 5 to Ravi, another paying all 5 to Meera. Half the network hears about the Ravi payment first; the other half hears about the Meera one first. Each looks valid on its own. Which one counts? Without a shared answer, the same coins get spent twice. This is the double-spend problem.
The rule
Consensus is how strangers agree on one order, and one order stops double spends.
Once every node agrees which transaction came first, the second is simply invalid, because those coins are already gone. So the real job of a consensus protocol is to agree on the order of blocks, and with it the order of transactions.
Three things make that hard on an open network:
- No boss. There is no central server entitled to announce "Ravi's came first".
- Faulty and dishonest nodes. Some nodes crash, lag or lie. A node that may fail in any way at all, including deliberate lying, is called Byzantine, after a classic puzzle about generals who cannot trust all of their messengers.
- Identities are free. On a public network anyone can start a thousand nodes. If agreement were "one node, one vote", an attacker would simply run more nodes. This is a Sybil attack.
The answer to the last problem shapes every design: blockchains tie influence to something scarce and costly to fake.
| Protocol family | What earns you a say | Covered in |
|---|---|---|
| Proof of work | Computing power spent on hashing | Lesson 2 |
| Proof of stake | Coins locked up as collateral | Lesson 3 |
| BFT voting | A seat in a known set of validators | Lesson 4 |
This is the real meaning of decentralization versus trust. You no longer trust a particular institution. Instead you trust that the rules, the maths and the cost of cheating keep enough participants honest.
The shortcut
Meet any consensus design with two questions: "Who gets to propose the next block?" and "What stops them from lying?"
Two conflicting transactions
"Asha pays Ravi 5" and "Asha pays Meera 5" spend the same coins. Nodes disagree about which arrived first.
The trap is thinking consensus establishes the "truth" about the outside world. It only settles the order and validity of transactions inside the chain.
Drills
Answer fast — trust the rule.
What problem does agreeing on one order of transactions solve?
Why can't a public blockchain decide by "one node, one vote"?
What does it mean to call a node Byzantine?
The network settles on a block containing "Asha pays Ravi 5". What happens to "Asha pays Meera 5"?
Remember
- Consensus is how strangers agree on one order, and one order stops double spends.
- Shortcut: ask who proposes the next block and what stops them from lying.
- Trap: influence must cost something scarce, such as work, stake or a known seat, or fake nodes win.