How Does Blockchain Work? Blocks, Nodes And Consensus

Imagine several people keeping copies of the same account book. A new payment arrives. Who checks it, which copy gets updated, and what stops someone quietly rewriting an old entry?

A blockchain tackles that problem with a shared digital ledger, a record of transactions maintained across a network. New records are grouped into blocks. Each block links to the one before it, while the network follows rules for accepting changes.

This is the technology underneath cryptocurrencies such as Bitcoin (BTC), but it isn't another name for Bitcoin. Blockchains can also support other applications, including registries and records shared between organisations. Different networks choose different rules and permissions.

What is being shared?

The ledger contains a history of accepted transactions. A node is a computer running software that participates in the network. Nodes communicate with one another, and those that maintain and check the ledger can build their own view of the accepted record.

The important difference from a company database is who maintains that record. With a blockchain, copies are held across network nodes. The design lets participants check updates against agreed rules rather than treating one operator's database as the sole answer.

Sharing copies alone isn't enough. If two copies disagree, the network needs a way to decide which history to follow. That is where validation and consensus come in.

How does a transaction reach the chain?

Take a simple transfer between two accounts. The exact format depends on the network, but Ethereum's transaction documentation provides a useful example of the steps.

1. Your wallet prepares an instruction, including the destination, amount and fee settings.
2. It signs the instruction using a private key, secret information that authorises actions from the account. Other computers can verify the digital signature without receiving that secret.
3. The signed transaction is broadcast to the network. It is a request at this point, not an accepted payment.
4. A participant proposes a block containing transactions. Other validating computers check the proposed transactions and their results against the network's rules.
5. The network's agreement process determines which valid blocks become the accepted history.

A valid signature answers whether the transaction was authorised by the relevant key. It doesn't answer every other question. The network also checks conditions such as available funds and whether an instruction is allowed by the program it calls.

The order matters too. A person cannot spend the same balance twice merely by sending two instructions at once. The accepted sequence determines which actions are valid against the balance left by earlier actions.

Why put transactions into blocks?

A block groups records so the network can agree on a batch and its order. It also contains a reference to the preceding block, usually through a cryptographic hash.

A hash is a digital fingerprint calculated from data. Change the data and the fingerprint changes. As Ethereum's blocks documentation explains, linking blocks through hashes makes changes to earlier records detectable: later blocks point back to data that would no longer match.

That creates the chain in “blockchain”. It doesn't make a record physically impossible to change. The network's validation and agreement rules are what make an altered history difficult to get accepted. NIST describes the result as tamper evident and tamper resistant, rather than treating the word “immutable” as magic.

How do computers agree without one boss?

Consensus is the process for agreeing on the network's accepted history. It includes rules for proposing blocks, checking them and choosing between competing valid histories. Different blockchains use different designs.

Bitcoin uses a system based on proof of work. Miners compete by performing computational work to propose blocks. Our Bitcoin mining explainer explains that competition and its rewards.

Ethereum (ETH) uses a system based on proof of stake. Its validators commit ETH as security and participate in proposing and confirming blocks. Rewards and penalties are part of the design. Ethereum's consensus documentation explains the difference between these approaches.

The ability to propose a block is not permission to ignore transaction rules. Other validating nodes check the proposal. A blockchain needs both a way to choose the history and rules for deciding what that history is allowed to contain.

Agreement also takes time. Two proposed blocks can briefly compete before the network settles on a history. Networks differ in when a transaction is treated as final, meaning it has reached the point their rules consider settled. A transaction appearing in a wallet and reaching that point need not be the same event.

Are all blockchains public and decentralised?

No. NIST's blockchain overview distinguishes permissionless networks, where participation in publishing blocks does not require an authority's approval, from permissioned networks, where only approved participants can do so.

Public and private describe access to the record. Permissioned and permissionless describe who is allowed to take part in particular network activities. These are related choices, but not identical ones. A network can make records publicly readable while restricting who validates them.

A group of organisations might use a permissioned chain to share records among approved participants. Its trust arrangement differs from an open cryptocurrency network. Calling both “blockchain” doesn't tell you how many independent parties control them or who can change their rules.

What can't a blockchain prove?

It can help establish which data was accepted and whether later changes conflict with that record. It cannot, by itself, prove that information entered from the outside world was true.

Suppose a record says a parcel was delivered. Keeping that statement on a blockchain doesn't establish that the parcel reached the door. The person or device supplying the information could be mistaken. NIST calls this the problem of reaching beyond the digital record.

The same issue affects applications using prices or weather reports. An oracle brings external information to a blockchain program. The program can execute exactly as written while relying on an inaccurate input.

Nor does a valid transaction prove that the user intended it. A stolen private key can authorise a transfer that passes the network's checks. Our wallet explainer explains why control of that key matters.

The record, the rules and the inputs answer different questions. A blockchain can make a shared history harder to rewrite. Whether that history describes a genuine delivery, a sound investment or a correctly written application still needs evidence beyond the chain.