Blockchain: Definition, How It Works, Types, History, Uses and Limitations

A blockchain is a shared digital ledger that can only be added to. Copies are kept on many computers (nodes), and the network usually runs without a central authority such as a bank, company or government. The U.S. National Institute of Standards and Technology (NIST) calls blockchains "tamper evident and tamper resistant digital ledgers." Under normal operation, a transaction cannot be changed once it has been published [1]. Records are grouped into blocks, and each block is cryptographically linked to the one before it. Nodes use a consensus mechanism to agree on which blocks to add.

What is a blockchain?

A blockchain is one type of distributed ledger technology (DLT). The OECD describes DLT as a set of technologies that together create "a digital, shared and self-updating ledger of verified transactions or information among parties in a network" . What sets a blockchain apart is its structure: records are packed into blocks, and the blocks form a single chain in time order. Several independent parties hold the ledger and check it against the same rules, instead of trusting one database operator.

Three properties come up in most definitions:

Append-only history: new data is added in new blocks. Old blocks are not edited, so the full transaction history stays available [1].
Cryptographic linking: each block stores the hash of the previous block's header. Changing an old transaction would therefore mean changing every block after it [5].
Distributed agreement: nodes check blocks independently against the same consensus rules. When they accept the same blocks, the network is said to be in consensus [5].

How a blockchain works

Blocks and hashes

In Bitcoin, the transactions in a block are hashed in pairs, over and over, until a single value is left, called the Merkle root. The block header stores this Merkle root along with the hash of the previous block's header [5]. Ethereum blocks likewise contain a reference to their parent block. Ethereum's documentation notes that "one change in any block in history would invalidate all the following blocks" [10]. This chaining is why a blockchain is described as tamper-evident: a change is easy to spot, even though the design does not physically prevent one.

Consensus mechanisms

A consensus mechanism decides which participant may add the next block and how the other participants check it. The main approaches differ in cost, speed and who is allowed to take part.

Mechanism	How the next block is chosen	Example and notable parameters
Proof of work (PoW)	Miners compete to find a block hash below a target value, which takes a lot of computation [5].	Bitcoin adjusts mining difficulty every 2,016 blocks to keep the average block time at about 10 minutes [5].
Proof of stake (PoS)	Validators lock up value that can be destroyed ("slashed") if they act dishonestly. A validator is picked at random to propose each block [9].	On Ethereum, a validator deposits 32 ETH. Time is divided into 12-second slots, and 32 slots make an epoch [9].
Ordering service (permissioned)	Known, identified participants order transactions with a pluggable protocol that can be crash- or Byzantine-fault-tolerant [11].	Hyperledger Fabric executes and endorses transactions first, then orders them, then validates them against an endorsement policy [11].

In Ethereum's proof of stake, a block becomes final when validators holding two-thirds of the staked ETH vote on checkpoint pairs across epoch boundaries. After that, reversing it would be extremely expensive [9].

Forks

A fork happens when a blockchain splits into two versions. Temporary forks occur when two blocks are found at about the same time. Bitcoin nodes then follow the chain that took the most work to build [5]. Rule changes can also cause forks. In a hard fork, nodes that have not upgraded reject blocks made under the new rules, which can split the network permanently. In a soft fork, upgraded nodes stay compatible with the old rules, which avoids a permanent split if they control enough of the network's hash rate [5].

Public vs private blockchains (permissionless and permissioned)

NIST defines a permissionless system as one where "all users' permissions are equal and not set by any administrator or consortium" [3]. In a permissioned system, every node and every user must be given permission to use it, usually by an administrator or a consortium [2].

Aspect	Public (permissionless)	Private or consortium (permissioned)
Who can participate	Almost anyone, often under a pseudonym [11]	Known, identified and often vetted participants [11]
Typical consensus	Proof of work or proof of stake, with native-token incentives [11]	Ordering protocols that need no mining [11]
Examples	Bitcoin, Ethereum	Hyperledger Fabric, Corda
Main trade-off	Open participation and public verifiability, but limited scalability	Better efficiency, but participants rely on the governance body that grants access [11]

History of blockchain technology

1991: Stuart Haber and W. Scott Stornetta published "How to Time-Stamp a Digital Document" in the Journal of Cryptology, volume 3. It describes hash-linked timestamps that are now seen as a forerunner of the blockchain. The Bitcoin white paper cites their work three times [7].
31 October 2008: Someone using the pseudonym Satoshi Nakamoto published the Bitcoin white paper [6].
3 January 2009: The Bitcoin genesis block was mined. The first transaction followed on 12 January 2009 [6].
1 December 2020: Ethereum's proof-of-stake Beacon Chain launched and ran alongside the proof-of-work main network [8].
15 September 2022: In the Merge, Ethereum switched to proof of stake and stopped mining [8].

Smart contracts

NIST defines a smart contract as "a collection of code and data (sometimes referred to as functions and state)" that is deployed on the blockchain network using cryptographically signed transactions [4]. The network's nodes run the code, every node must get the same result, and the result is recorded on the blockchain [4]. Smart contracts let a blockchain record application logic, such as token issuance or escrow rules, and not just simple transfers of value.

Uses of blockchain

The OECD lists these application areas for blockchain and other distributed ledgers :

Cryptocurrencies and other financial services, including asset tokenisation
Digital identity systems
Public-sector administration and record keeping
Supply chain traceability
Automated agreements through smart contracts

A 2017 academic analysis looked at supply chain management, interbank and international payments, and decentralised autonomous organisations. It concluded that a blockchain is not the right technical solution for every problem .

Limitations, risks and misconceptions

Immutability has limits. Tamper resistance depends on honest participants holding most of the network's power. In proof of work, a "51% attack" means controlling most of the hashing power [5]. In proof of stake, it means controlling at least half of the staked ETH. Ethereum's documentation adds that the honest community can respond, for example by forcibly removing the attacker's stake [9].
Accuracy of inputs. A blockchain records what it is given. It cannot confirm that off-chain data or assets linked to tokens are real. The OECD lists custody of access credentials, privacy, security weaknesses and cryptographic weaknesses among the risks .
Key management. Access depends on cryptographic credentials. Losing or exposing them is a recognised risk .
Scalability. Larger blocks need more computing power to process within each slot, which can push the network toward centralisation [10].
Energy use. Proof of work uses a lot of energy. The University of Cambridge's revised index put Bitcoin's electricity use in 2022 at 95.5 TWh [12]. The U.S. Energy Information Administration estimated that cryptocurrency mining made up 0.6% to 2.3% of U.S. electricity consumption . Ethereum's move to proof of stake cut its energy consumption by an estimated 99.95% [8].

When does a blockchain make sense?

Wüst and Gervais compared permissionless blockchains, permissioned blockchains and ordinary centralised databases. They offer a structured method for deciding which one fits, and argue against using a blockchain by default . Questions often asked in this kind of assessment include:

Do several parties need to write to a shared record?
Do those parties distrust each other, or distrust any single operator?
Could a trusted third party that is always online do the job instead?
Are the writers known and identifiable? If so, a permissioned design may be enough.
Does the record need to be publicly verifiable, or can access be restricted?

If one trusted party can keep the record, a conventional database is usually simpler and more efficient. A blockchain adds the most value when parties that do not fully trust each other need a shared history they can check themselves .