ONE SUBJECT · THREE WAYS IN
Understand dApps at your own pace
What is an application? What does the “d” change? Why decentralise a service? Start with these foundations, then choose how far you want to go.
Free · No wallet connection · 2 explained situations per level
Beginner · Applications, decentralisation and its purpose: foundations before examples.
First, what is an application?
An application is software designed to provide a service. It receives information, follows instructions and displays a result. You can use it on a phone, on a computer or in a browser. A dApp is also an application: the “d” describes how part of that service is organised.
1. An application is more than what you see
Buttons, menus and screens make up its interface. Behind them, code processes requests and data is stored. Everything may run on your device. But many online services also use servers: remote computers that respond to users’ requests.
When we compare a dApp with a “traditional” application here, we mean this second case: an online service whose essential programs and data are managed by an operator. Not every ordinary application follows this model.
2. Centralised: an operator holds control
In a centralised application, a company or organisation administers the service: it manages the servers, database and program that applies the rules. Users depend on this operator for that part of the service to work.
“Centralised” does not mean “one computer”. A company can use thousands of servers across several countries while retaining sole control. Having more machines does not automatically distribute decision-making power.
3. The “d” stands for “decentralized”
dApp is short for “decentralized application”. The idea is to distribute some functions and their control among multiple participants rather than entrusting them entirely to a single operator.
In the blockchain dApps discussed here, important service rules are executed and verified by a network. A blockchain maintains a shared record: different participants can check operations against the same rules. A smart contract is a program on that network that applies rules. It is not a contract that thinks or makes decisions on its own.
The screen can look almost identical to a traditional application. The change is behind it: for the operations concerned, you rely on the network’s rules and shared program, rather than asking only one company’s server to decide and record the outcome.
4. Traditional app or dApp: the differences
Compare a centralised online application with a dApp whose important functions rely on a public blockchain.
- Who applies the rules?
- Traditional application: a program managed by the operator on its infrastructure. dApp: for the on-chain part, programs executed and verified under the network’s rules.
- Who maintains the essential data?
- Traditional application: the operator administers its database. dApp: on-chain data is kept in a shared record checked by network participants. Other data may still live on ordinary servers.
- Who can change how it works?
- Traditional application: the team controlling the service can update its programs and database. dApp: it depends on the permissions provided. Some rules are fixed; others can be changed by administrators or a vote. Check the particular project.
5. What is the point of decentralising?
The aim is to reduce dependence on one party for certain operations. This can be useful when several people or organisations want to use common rules and data without giving one of them exclusive control over the whole system.
- Being able to verify
- On a public blockchain, recorded operations can be examined. A user or another program can check an outcome instead of relying solely on what the website displays.
- Sharing rules across applications
- Different interfaces can use the same programs accessible on the network. The service does not necessarily have to be tied to a single website.
- Reducing dependence on one access point
- If a website shuts down, on-chain programs may remain usable through another interface, provided the network operates and the contracts remain accessible. This does not guarantee that the entire dApp will continue without interruption.
6. Why not decentralise everything?
Coordinating a network adds constraints: possible fees, delays and harder error correction. Data placed on a public blockchain is visible; it is not a private vault for personal information.
A traditional application may be easier to maintain, faster and better suited to private data. Decentralisation is an organisational choice that suits some needs, not an automatic mark of quality.
A dApp can still depend on a website, data provider or administrators. It can have bugs. “Decentralised” therefore means neither “nobody is responsible”, “impossible to block” nor “risk-free”. The useful question is: which functions are actually decentralised, and who retains which powers?
Now an example: a directory of names with ENS
A directory links a name to an address. A traditional application could provide this service by keeping the associations in its operator’s database. ENS, a real naming system, relies on Ethereum and its programs to manage these associations. Different compatible applications can look them up. The on-screen result may look similar: the difference lies in how names and permissions are managed behind it.
- The service
Find an address from a readable name.
- The shared part
ENS uses data and rules on Ethereum.
- The interfaces
ENS Manager and other compatible applications use this system.
The directory analogy explains this particular use. Not all dApps are directories. Looking up a name transfers no money and does not certify its owner’s identity.
Check what you understood
Read the clues and explain your choice. Each answer has its own explanation; you can try again or read the full solution directly. These situations are fictional.
SITUATION 1 / 2
A company describes its application as decentralised.
- It uses 100 servers spread across several countries.
- It alone controls those servers, the program and the database.
Is that enough to call the service decentralised?
Read the full explanation
Answer: No: control remains with a single operator.
- Yes: many servers mean decentralisation.
The number of machines describes distributed infrastructure. Here, control remains concentrated in one company.
- No: control remains with a single operator.
Correct. The key question is who controls and verifies the operation, not just where the computers are located.
- Yes: having many users is enough.
The number of users does not automatically give them control over the service’s rules or data.
Remember: Geographically distributed infrastructure can still be centrally controlled.
SITUATION 2 / 2
Several organisations want to use a shared record.
- They want to be able to verify operations.
- None wants another to administer the record alone.
- They must also examine costs and data confidentiality.
What could be the benefit of a decentralised solution here?
Read the full explanation
Answer: Sharing verification without giving one organisation all the control.
- Guaranteeing that all data will be private.
A public blockchain does not guarantee confidentiality. That requirement must be considered separately before choosing an architecture.
- Removing every cost and risk.
Network coordination, mistakes and dependencies still matter. Decentralisation does not make them disappear.
- Sharing verification without giving one organisation all the control.
Correct. That is the need to examine here. They must then check that the solution really distributes control and meets their cost and confidentiality requirements.
Remember: Choose a decentralised structure to meet a need for shared control, not just because the word sounds modern.
Useful first: understand the difference between centralised and decentralised applications in the very simple version.
Reading information and changing it are different journeys
A dApp lets you read data and may also let you change it. These actions follow different journeys. For the on-chain part, a smart contract is the program that defines rules, including which changes are allowed. Let’s look at the roles of the interface, wallet and network.
Reading: asking the network a question
An application can ask “which address is associated with this name?” without publishing a transaction. The read does not change the record and requires no transaction fee. A service may still charge for access to its data.
Writing: executing an authorised request
In our scenario, the transaction asks the contract to record B. The contract checks the account’s permissions. Clicking or signing does not prove the update is complete: the request may still be pending or fail.
An executed transaction uses resources measured in gas. It can incur fees even when execution fails. The amount depends on the operation and network; we do not invent a fixed price here.
Three requests that should not be confused
Wallet windows can look similar. Their effects depend on what they actually request.
- Connection
- Presenting the selected account to the website. This alone does not grant a new permission to spend tokens.
- Message signature
- For example, proving control of an account to log into a service. Other messages can authorise spending: “no fee” does not mean “no effect”.
- Transaction
- Requesting an action on the network. It might update data, transfer assets or grant a permission; it is not necessarily a money transfer.
What about swapping tokens with Uniswap?
That is another use. A token is an asset represented on a blockchain. A swap may require permission for a contract to spend that token, followed by the swap itself. An approval is not proof that the swap took place. The steps vary between flows.
This spending mechanism is not the definition of a dApp, and it is not needed simply to read an ENS address. Our permissions workshop covers it separately.
Lea wants to change the address linked to her name
Suppose Lea controls an ENS name with a record stored on Ethereum. She replaces address A with address B. This is a simplified on-chain update scenario: the exact screens and permissions depend on the name and contract involved.
- Prepare
The interface prepares a request to change the record to B.
- Check and sign
The wallet presents the request: network, action, contract and fees.
- Execute and verify
The transaction must be included and succeed. Lea then reads the record again.
A and B are fictional labels, not usable addresses. Changing this association does not transfer funds already held at A.
Check what you understood
Read the clues and explain your choice. Each answer has its own explanation; you can try again or read the full solution directly. These situations are fictional.
SITUATION 1 / 2
Lea has signed an update to the address linked to her ENS name.
- Transaction status: pending.
- The current lookup still shows address A.
Can she announce that address B is already recorded?
Read the full explanation
Answer: No: wait for the outcome and check the record.
- No: wait for the outcome and check the record.
Correct. The signature authorises a request; a pending status does not prove execution.
- Yes: signing is enough to change the data.
Signing does not replace execution. The transaction may remain pending or fail.
- Yes: disconnecting will force the update.
Disconnecting the website does not force the network to execute a transaction.
Remember: Even after a reported success, read the expected information on the correct network.
SITUATION 2 / 2
On a swap dApp, Lea wanted to exchange 25 units of a token.
- An approval to spend 25 units succeeded.
- No swap operation has been executed yet.
Which conclusion is justified?
Read the full explanation
Answer: The permission exists; the swap is a separate action.
- She must have received the other token.
A successful permission does not prove that a swap executed. The scenario states that no swap has taken place.
- The 25 units are network fees.
Here, 25 is the spending limit granted to the contract, not the cost of the transaction.
- The permission exists; the swap is a separate action.
Correct. She still needs to check the swap request and its outcome. This permission should not be confused with fees.
Remember: A permission describes what a contract may do; it does not prove that the contract has done it.
Prerequisites: accounts, signatures, transactions and smart contracts. This technical explanation is not a security audit.
Trace the data back to its source
A dApp combines several components with different dependencies. Identify who provides the interface, who responds to reads, which contracts execute operations and who can change their parameters. Decentralisation is not an all-or-nothing property.
An RPC read is not a published transaction
Here, RPC is the interface used to query an Ethereum node. The eth_call method executes a call in a given context without creating an on-chain transaction. Its result depends on factors including the network, contract, parameters and requested block.
A successful simulation does not promise future execution: state can change between the read and a transaction’s inclusion. Comparing two results therefore requires specifying the same context, including the block. An RPC provider remains a dependency to examine.
Verify a write and its intended effect
For Lea’s fictional update, distinguish signing, broadcasting, inclusion in a block and finality. The execution receipt gives a status. Also check that the correct operation targeted the correct contract, then read the expected record again.
A transaction hash identifies the request, not its success. A successful receipt for a token approval does not prove that a swap happened. An interface displaying old data might have a stale cache: compare networks and blocks before concluding that the contract is faulty.
A displayed name also needs checking
Reverse resolution starts from an address to find a name. ENS recommends following it with a forward lookup of that name and checking that the resulting address matches the starting address. Otherwise, display the address instead of the name. Technical consistency still does not prove a person’s legal identity.
Where do dependencies remain?
A technical review should at least ask these questions; this list does not replace an audit.
- Interface and data
- Who hosts the website? Which RPC or indexer responds? Does the displayed data match the expected network and block?
- Contracts and control
- Which permissions allow parameters to change, an implementation to be replaced or a function to be paused? Check the specific contract; do not assume every dApp has these powers.
- External sources
- Does the service rely on an oracle (an external data source), a gateway or an operator? Using contracts does not remove those dependencies.
ENS: the registry and the resolver do different jobs
In the classic model of on-chain ENS resolution on Ethereum, the registry identifies the resolver responsible for a name. The resolver is the contract queried for a record, such as an address. The registry does not itself hold every piece of profile data.
- Registry
Which resolver is associated with this name?
- Resolver
What address record does it return?
- Application
How does it verify and display that response?
A conceptual diagram, not an implementation to copy. Libraries also handle mechanisms such as wildcard resolution and off-chain data; not every name follows two identical direct calls.
Check what you understood
Read the clues and explain your choice. Each answer has its own explanation; you can try again or read the full solution directly. These situations are fictional.
SITUATION 1 / 2
You are investigating a display after an update. These observations are fictional.
- Same network, name and resolver.
- Read at block 100: address A.
- Successful update in block 101.
- Read at block 102: address B. The interface still shows A from its block-100 cache.
Which diagnosis do these observations support?
Read the full explanation
Answer: The interface displays data from before the update.
- The network gives two answers for the same state.
The reads concern different blocks, before and after a write. They do not describe the same state.
- The interface displays data from before the update.
Correct. The block-100 cache explains A; the block-102 read gives B. Refresh the display with the correct context.
- A second transaction must be signed immediately.
The write already succeeded in this scenario. A new signature does not fix the interface’s cache.
Remember: Before comparing reads, compare their context: network, contract, parameters and block.
SITUATION 2 / 2
Your application wants to display a name instead of address A. The letters are fictional labels.
- Reverse resolution of A suggests a name.
- Forward resolution of that name returns address B.
- A and B are different.
What should the application display here?
Read the full explanation
Answer: Address A, without replacing it with that name.
- Address A, without replacing it with that name.
Correct. The round-trip check failed: the name does not match the starting address.
- The name, because the first lookup returned it.
Reverse resolution alone is insufficient. The forward check must return the original address.
- Address B as account A’s new identity.
This result does not turn A into B. It indicates inconsistent resolution, not an identity change.
Remember: A candidate name must be checked both ways; this does not certify the person behind the account.
Other questions about dApps
What does dApps mean?
dApp stands for “decentralized application”. dApps is simply the plural. This guide uses Ethereum-based applications as examples; dApps can also run on other networks.
Do I have to buy crypto to understand dApps?
No. This guide and its exercises are free, with no account or wallet connection. Using a real service is a separate decision: transactions can incur fees and financial operations can cause losses.
Does every dApp need its own token?
No. Having a token with an application’s name is not what makes it a dApp. Ask which part of its operation relies on a decentralised network. Network fees and any token used by a service are separate questions.
Does disconnecting remove spending permissions?
Not necessarily. Disconnecting the website and revoking an existing token spending approval are different operations. The permissions workshop explains the distinction.
Sources and further reading
- Ethereum.org — Introduction to dapps · in English
- Uniswap — How the protocol works · in English
- ENS — Protocol and ENS Manager · in English
- MDN — Client-server overview · in English
- Ethereum.org — Smart contracts · in English
- Ethereum.org — Nodes and clients · in English
- MetaMask — Connecting to a dapp · in English
- MetaMask — Token approvals · in English
- MetaMask — Disconnecting and revoking permissions · in English
- ENS — Resolution and reverse verification · in English
- Ethereum.org — JSON-RPC: eth_call and transaction receipts · in English
- Ethereum.org — Transactions · in English
- Ethereum.org — Gas and fees · in English
- ERC-2612 — Signed token approvals · in English
- Ethereum.org — Upgrading smart contracts · in English
- Ethereum.org — Oracles · in English
Page updated on
Documentation checked on · WhyTheBlockchain · Editorial approach