In this lesson you will learn to:
- Tell the difference between centralised, decentralised and distributed systems.
- Explain the Byzantine Generals problem in one paragraph.
- Describe why agreement is hard when some participants may lie.
Three shapes of a network
Imagine drawing every participant in a system as a dot and every connection as a line.
- In a centralised system, every dot connects to one central dot. The centre keeps the record and makes the decisions. Most online services work this way.
- In a decentralised system, there are several hubs, each serving its own group. No single hub controls everything, but each group depends on its hub.
- In a distributed system, dots connect to many other dots, and there is no hub at all. Every participant can, in principle, check everything.
People often use "decentralised" loosely to mean "not centralised". In this course, what matters is the practical question: who can change the record, and who can check it?
Why distributed agreement is hard
In a centralised system, agreement is easy: the centre decides. In a distributed system, every participant has its own copy, messages take time to travel, some computers go offline, and some participants may be dishonest. Somehow they must all end up with the same record.
The Byzantine Generals problem
In 1982 the computer scientists Leslie Lamport, Robert Shostak and Marshall Pease described this challenge as a story. Several generals surround a city. They must all attack together or all retreat together; a split decision means defeat. They can only communicate by messenger, and some generals may be traitors who send different messages to different colleagues.
The question is: can the loyal generals always reach the same decision? The paper showed that with traditional messages, agreement is possible only if fewer than one-third of the generals are traitors — and only if everyone knows who the participants are.
The open-membership twist
Classic solutions assume a known list of participants. A public blockchain is harder: anyone can join, anonymously, at any time. One person could create thousands of fake identities and outvote everyone else. This is called a Sybil attack, after a book about a person with many identities.
The key insight of public blockchains is to make influence over the record depend on something that is costly to fake — computing work in proof of work, or locked-up tokens in proof of stake — rather than on the number of identities. Lesson 3.3 explains both methods.
Not every blockchain is open
Some organisations run permissioned blockchains, where only approved participants can keep copies or add records. These can use classic agreement methods because members are known. They trade openness for speed and control. Most of this course focuses on open, public blockchains, because they are the harder and more unusual design.
Key takeaways
- Centralised systems have one keeper; distributed systems have none.
- Distributed agreement is hard because messages are slow and some participants may lie.
- The Byzantine Generals problem formalised this difficulty in 1982.
- Public blockchains tie influence to a costly resource to resist fake identities.
Check yourself
In a centralised system, who decides what goes into the record?
The central organisation that keeps the authoritative copy.
What is a Sybil attack?
Creating many fake identities to gain outsized influence in a system where anyone can join.
How do public blockchains resist Sybil attacks?
By tying influence to something costly to fake — computing work or locked-up tokens — rather than to the number of identities.
Sources
This lesson is for general education about how the technology works. It is not a recommendation to acquire, sell or hold any crypto-asset and is not financial advice.