'eth' is a protocol on the RLPx transport that facilitates exchange of Ethereum blockchain information between peers. The current protocol version is eth/66. See end of document for a list of changes in past protocol versions.
Once a connection is established, a Status message must be sent. Following the reception of the peer's Status message, the Ethereum session is active and any other message may be sent.
Within a session, three high-level tasks can be performed: chain synchronization, block propagation and transaction exchange. These tasks use disjoint sets of protocol messages and clients typically perform them as concurrent activities on all peer connections.
Client implementations should enforce limits on protocol message sizes. The underlying RLPx transport limits the size of a single message to 16.7 MiB. The practical limits for the eth protocol are lower, typically 10 MiB. If a received message is larger than the limit, the peer should be disconnected.
In addition to the hard limit on received messages, clients should also impose 'soft' limits on the requests and responses which they send. The recommended soft limit varies per message type. Limiting requests and responses ensures that concurrent activity, e.g. block synchronization and transaction exchange work smoothly over the same peer connection.
Nodes participating in the eth protocol are expected to have knowledge of the complete chain of all blocks from the genesis block to current, latest block. The chain is obtained by downloading it from other peers.
Upon connection, both peers send their Status message, which includes the Total Difficulty (TD) and hash of their 'best' known block.
The client with the worst TD then proceeds to download block headers using the GetBlockHeaders message. It verifies proof-of-work values in received headers and fetches block bodies using the GetBlockBodies message. Received blocks are executed using the Ethereum Virtual Machine, recreating the state tree and receipts.
Note that header downloads, block body downloads and block execution may happen concurrently.
Protocol versions eth/63 and later also allow synchronizing transaction execution results (i.e. state tree and receipts). This may be faster than re-executing all historical transactions but comes at the expense of some security.
State synchronization typically proceeds by downloading the chain of block headers, verifying their validity. Block bodies are requested as in the Chain Synchronization section but block transactions aren't executed, only their 'data validity' is verified. The client picks a block near the head of the chain and downloads merkle tree nodes and contract code incrementally by requesting the root node, its children, grandchildren, ... using GetNodeData until the entire tree is synchronized.
Newly-mined blocks must be relayed to all nodes. This happens through block propagation, which is a two step process. When a NewBlock announcement message is received from a peer, the client first verifies the basic header validity of the block, checking whether the proof-of-work value is valid. It then sends the block to a small fraction of connected peers (usually the square root of the total number of peers) using the NewBlock message.
After the header validity check, the client imports the block into its local chain by
executing all transactions contained in the block, computing the block's 'post state'. The
block's state-root
hash must match the computed post state root. Once the block is fully
processed, and considered valid, the client sends a NewBlockHashes message about the
block to all peers which it didn't notify earlier. Those peers may request the full block
later if they fail to receive it via NewBlock from anyone else.
A node should never send a block announcement back to a peer which previously announced the same block. This is usually achieved by remembering a large set of block hashes recently relayed to or from each peer.
The reception of a block announcement may also trigger chain synchronization if the block is not the immediate successor of the client's current latest block.
All nodes must exchange pending transactions in order to relay them to miners, which will pick them for inclusion into the blockchain. Client implementations keep track of the set of pending transactions in the 'transaction pool'. The pool is subject to client-specific limits and can contain many (i.e. thousands of) transactions.
When a new peer connection is established, the transaction pools on both sides need to be synchronized. Initially, both ends should send NewPooledTransactionHashes messages containing all transaction hashes in the local pool to start the exchange.
On receipt of a NewPooledTransactionHashes announcement, the client filters the received set, collecting transaction hashes which it doesn't yet have in its own local pool. It can then request the transactions using the GetPooledTransactions message.
When new transactions appear in the client's pool, it should propagate them to the network using the Transactions and NewPooledTransactionHashes messages. The Transactions message relays complete transaction objects and is typically sent to a small, random fraction of connected peers. All other peers receive a notification of the transaction hash and can request the complete transaction object if it is unknown to them. The dissemination of complete transactions to a fraction of peers usually ensures that all nodes receive the transaction and won't need to request it.
A node should never send a transaction back to a peer that it can determine already knows of it (either because it was previously sent or because it was informed from this peer originally). This is usually achieved by remembering a set of transaction hashes recently relayed by the peer.
Transaction objects exchanged by peers have one of two encodings. In definitions across
this specification, we refer to transactions of either encoding using the identifier txₙ
.
tx = {legacy-tx, typed-tx}
Untyped, legacy transactions are given as an RLP list.
legacy-tx = [
nonce: P,
gas-price: P,
gas-limit: P,
recipient: {B_0, B_20},
value: P,
data: B,
V: P,
R: P,
S: P,
]
EIP-2718 typed transactions are encoded as RLP byte arrays where the first byte is the
transaction type (tx-type
) and the remaining bytes are opaque type-specific data.
typed-tx = tx-type || tx-data
Transactions must be validated when they are received. Validity depends on the Ethereum chain state. The specific kind of validity this specification is concerned with is not whether the transaction can be executed successfully by the EVM, but only whether it is acceptable for temporary storage in the local pool and for exchange with other peers.
Transactions are validated according to the rules below. While the encoding of typed
transactions is opaque, it is assumed that their tx-data
provides values for nonce
,
gas-price
, gas-limit
, and that the sender account of the transaction can be determined
from their signature.
- If the transaction is typed, the
tx-type
must be known to the implementation. Defined transaction types may be considered valid even before they become acceptable for inclusion in a block. Implementations should disconnect peers sending transactions of unknown type. - The signature must be valid according to the signature schemes supported by the chain. For typed transactions, signature handling is defined by the EIP introducing the type. For legacy transactions, the two schemes in active use are the basic 'Homestead' scheme and the EIP-155 scheme.
- The
gas-limit
must cover the 'intrinsic gas' of the transaction. - The sender account of the transaction, which is derived from the signature, must have
sufficient ether balance to cover the cost (
gas-limit * gas-price + value
) of the transaction. - The
nonce
of the transaction must be equal or greater than the current nonce of the sender account. - When considering the transaction for inclusion in the local pool, it is up to implementations to determine how many 'future' transactions with nonce greater than the current account nonce are valid, and to which degree 'nonce gaps' are acceptable.
Implementations may enforce other validation rules for transactions. For example, it is common practice to reject encoded transactions larger than 128 kB.
Unless noted otherwise, implementations must not disconnect peers for sending invalid transactions, and should simply discard them instead. This is because the peer might be operating under slightly different validation rules.
Ethereum blocks are encoded as follows:
block = [header, transactions, ommers]
transactions = [tx₁, tx₂, ...]
ommers = [header₁, header₂, ...]
header = [
parent-hash: B_32,
ommers-hash: B_32,
coinbase: B_20,
state-root: B_32,
txs-root: B_32,
receipts-root: B_32,
bloom: B_256,
difficulty: P,
number: P,
gas-limit: P,
gas-used: P,
time: P,
extradata: B,
mix-digest: B_32,
block-nonce: B_8
]
In certain protocol messages, the transaction and ommer lists are relayed together as a single item called the 'block body'.
block-body = [transactions, ommers]
The validity of block headers depends on the context in which they are used. For a single
block header, only the validity of the proof-of-work seal (mix-digest
, block-nonce
)
can be verified. When a header is used to extend the client's local chain, or multiple
headers are processed in sequence during chain synchronization, the following rules apply:
- Headers must form a chain where block numbers are consecutive and the
parent-hash
of each header matches the hash of the preceding header. - When extending the locally-stored chain, implementations must also verify that the
values of
difficulty
,gas-limit
andtime
are within the bounds of protocol rules given in the Yellow Paper. - The
gas-used
field of a block header must be less than or equal to thegas-limit
.
For complete blocks, we distinguish between the validity of the block's EVM state transition, and the (weaker) 'data validity' of the block. The definition of state transition rules is not dealt with in this specification. We require data validity of the block for the purposes of immediate block propagation and during state synchronization.
To determine the data validity of a block, use the rules below. Implementations should disconnect peers sending invalid blocks.
- The block
header
must be valid. - The
transactions
contained in the block must be valid for inclusion into the chain at the block's number. This means that, in addition to the transaction validation rules given earlier, validating whether thetx-type
is permitted at the block number is required, and validation of transaction gas must take the block number into account. - The sum of the
gas-limit
s of all transactions must not exceed thegas-limit
of the block. - The
transactions
of the block must be verified against thetxs-root
by computing and comparing the merkle trie hash of the transactions list. - The
ommers
list may contain at most two headers. keccak256(ommers)
must match theommers-hash
of the block header.- The headers contained in the
ommers
list must be valid headers. Their block number must not be greater than that of the block they are included in. The parent hash of an ommer header must refer to an ancestor of depth 7 or less of its including block, and it must not have been included in any earlier block contained in this ancestor set.
Receipts are the output of the EVM state transition of a block. Like transactions,
receipts have two distinct encodings and we will refer to either encoding using the
identifier receiptₙ
.
receipt = {legacy-receipt, typed-receipt}
Untyped, legacy receipts are encoded as follows:
legacy-receipt = [
post-state-or-status: {B_32, {0, 1}},
cumulative-gas: P,
bloom: B_256,
logs: [log₁, log₂, ...]
]
log = [
contract-address: B_20,
topics: [topic₁: B, topic₂: B, ...],
data: B
]
EIP-2718 typed receipts are encoded as RLP byte arrays where the first byte gives the
receipt type (matching tx-type
) and the remaining bytes are opaque data specific to the
type.
typed-receipt = tx-type || receipt-data
In the Ethereum Wire Protocol, receipts are always transferred as the complete list of all
receipts contained in a block. It is also assumed that the block containing the receipts
is valid and known. When a list of block receipts is received by a peer, it must be
verified by computing and comparing the merkle trie hash of the list against the
receipts-root
of the block. Since the valid list of receipts is determined by the EVM
state transition, it is not necessary to define any further validity rules for receipts in
this specification.
In most messages, the first element of the message data list is the request-id
. For
requests, this is a 64-bit integer value chosen by the requesting peer. The responding
peer must mirror the value in the request-id
element of the response message.
[version: P, networkid: P, td: P, blockhash: B_32, genesis: B_32, forkid]
Inform a peer of its current state. This message should be sent just after the connection is established and prior to any other eth protocol messages.
version
: the current protocol versionnetworkid
: integer identifying the blockchain, see table belowtd
: total difficulty of the best chain. Integer, as found in block header.blockhash
: the hash of the best (i.e. highest TD) known blockgenesis
: the hash of the genesis blockforkid
: An EIP-2124 fork identifier, encoded as[fork-hash, fork-next]
.
This table lists common Network IDs and their corresponding networks. Other IDs exist which aren't listed, i.e. clients should not require that any particular network ID is used. Note that the Network ID may or may not correspond with the EIP-155 Chain ID used for transaction replay prevention.
ID | chain |
---|---|
0 | Olympic (disused) |
1 | Frontier (now mainnet) |
2 | Morden (disused) |
3 | Ropsten (current PoW testnet) |
4 | Rinkeby |
For a community curated list of chain IDs, see https://chainid.network.
[[blockhash₁: B_32, number₁: P], [blockhash₂: B_32, number₂: P], ...]
Specify one or more new blocks which have appeared on the network. To be maximally helpful, nodes should inform peers of all blocks that they may not be aware of. Including hashes that the sending peer could reasonably be considered to know (due to the fact they were previously informed of because that node has itself advertised knowledge of the hashes through NewBlockHashes) is considered bad form, and may reduce the reputation of the sending node. Including hashes that the sending node later refuses to honour with a proceeding GetBlockHeaders message is considered bad form, and may reduce the reputation of the sending node.
[tx₁, tx₂, ...]
Specify transactions that the peer should make sure is included on its transaction queue. The items in the list are transactions in the format described in the main Ethereum specification. Transactions messages must contain at least one (new) transaction, empty Transactions messages are discouraged and may lead to disconnection.
Nodes must not resend the same transaction to a peer in the same session and must not relay transactions to a peer they received that transaction from. In practice this is often implemented by keeping a per-peer bloom filter or set of transaction hashes which have already been sent or received.
[request-id: P, [startblock: {P, B_32}, limit: P, skip: P, reverse: {0, 1}]]
Require peer to return a BlockHeaders message. The response must contain a number of block
headers, of rising number when reverse
is 0
, falling when 1
, skip
blocks apart,
beginning at block startblock
(denoted by either number or hash) in the canonical chain,
and with at most limit
items.
[request-id: P, [header₁, header₂, ...]]
This is the response to GetBlockHeaders, containing the requested headers. The header list may be empty if none of the requested block headers were found. The number of headers that can be requested in a single message may be subject to implementation-defined limits.
The recommended soft limit for BlockHeaders responses is 2 MiB.
[request-id: P, [blockhash₁: B_32, blockhash₂: B_32, ...]]
This message requests block body data by hash. The number of blocks that can be requested in a single message may be subject to implementation-defined limits.
[request-id: P, [block-body₁, block-body₂, ...]]
This is the response to GetBlockBodies. The items in the list contain the body data of the requested blocks. The list may be empty if none of the requested blocks were available.
The recommended soft limit for BlockBodies responses is 2 MiB.
[block, td: P]
Specify a single complete block that the peer should know about. td
is the total
difficulty of the block, i.e. the sum of all block difficulties up to and including this
block.
[txhash₁: B_32, txhash₂: B_32, ...]
This message announces one or more transactions that have appeared in the network and which have not yet been included in a block. To be maximally helpful, nodes should inform peers of all transactions that they may not be aware of.
The recommended soft limit for this message is 4096 hashes (128 KiB).
Nodes should only announce hashes of transactions that the remote peer could reasonably be considered not to know, but it is better to return more transactions than to have a nonce gap in the pool.
[request-id: P, [txhash₁: B_32, txhash₂: B_32, ...]]
This message requests transactions from the recipient's transaction pool by hash.
The recommended soft limit for GetPooledTransactions requests is 256 hashes (8 KiB). The recipient may enforce an arbitrary limit on the response (size or serving time), which must not be considered a protocol violation.
[request-id: P, [tx₁, tx₂...]]
This is the response to GetPooledTransactions, returning the requested transactions from the local pool. The items in the list are transactions in the format described in the main Ethereum specification.
The transactions must be in same order as in the request, but it is OK to skip transactions which are not available. This way, if the response size limit is reached, requesters will know which hashes to request again (everything starting from the last returned transaction) and which to assume unavailable (all gaps before the last returned transaction).
It is permissible to first announce a transaction via NewPooledTransactionHashes, but then to refuse serving it via PooledTransactions. This situation can arise when the transaction is included in a block (and removed from the pool) in between the announcement and the request.
A peer may respond with an empty list iff none of the hashes match transactions in its pool.
[request-id: P, [hash₁: B_32, hash₂: B_32, ...]]
Require peer to return a NodeData message containing state tree nodes or contract code matching the requested hashes.
[request-id: P, [value₁: B, value₂: B, ...]]
Provide a set of state tree nodes or contract code blobs which correspond to previously requested hashes from GetNodeData. Does not need to contain all; best effort is fine. This message may be an empty list if the peer doesn't know about any of the previously requested hashes. The number of items that can be requested in a single message may be subject to implementation-defined limits.
The recommended soft limit for NodeData responses is 2 MiB.
[request-id: P, [blockhash₁: B_32, blockhash₂: B_32, ...]]
Require peer to return a Receipts message containing the receipts of the given block hashes. The number of receipts that can be requested in a single message may be subject to implementation-defined limits.
[request-id: P, [[receipt₁, receipt₂], ...]]
This is the response to GetReceipts, providing the requested block receipts. Each element in the response list corresponds to a block hash of the GetReceipts request, and must contain the complete list of receipts of the block.
The recommended soft limit for Receipts responses is 2 MiB.
eth/66 (EIP-2481, April 2021)
Version 66 added the request-id
element in messages GetBlockHeaders, BlockHeaders,
GetBlockBodies, BlockBodies, GetPooledTransactions, PooledTransactions,
GetNodeData, NodeData, GetReceipts, Receipts.
eth/65 with typed transactions (EIP-2976, April 2021)
When typed transactions were introduced by EIP-2718, client implementers decided to accept the new transaction and receipt formats in the wire protocol without increasing the protocol version. This specification update also added definitions for the encoding of all consensus objects instead of referring to the Yellow Paper.
eth/65 (EIP-2464, January 2020)
Version 65 improved transaction exchange, introducing three additional messages: NewPooledTransactionHashes, GetPooledTransactions, and PooledTransactions.
Prior to version 65, peers always exchanged complete transaction objects. As activity and transaction sizes increased on the Ethereum mainnet, the network bandwidth used for transaction exchange became a significant burden on node operators. The update reduced the required bandwidth by adopting a two-tier transaction broadcast system similar to block propagation.
eth/64 (EIP-2364, November 2019)
Version 64 changed the Status message to include the EIP-2124 ForkID. This allows peers to determine mutual compatibility of chain execution rules without synchronizing the blockchain.
Version 63 added the GetNodeData, NodeData, GetReceipts and Receipts messages which allow synchronizing transaction execution results.
In version 62, the NewBlockHashes message was extended to include block numbers alongside the announced hashes. The block number in Status was removed. Messages GetBlockHashes (0x03), BlockHashes (0x04), GetBlocks (0x05) and Blocks (0x06) were replaced by messages that fetch block headers and bodies. The BlockHashesFromNumber (0x08) message was removed.
Previous encodings of the reassigned/removed message codes were:
- GetBlockHashes (0x03):
[hash: B_32, max-blocks: P]
- BlockHashes (0x04):
[hash₁: B_32, hash₂: B_32, ...]
- GetBlocks (0x05):
[hash₁: B_32, hash₂: B_32, ...]
- Blocks (0x06):
[[header, transactions, ommers], ...]
- BlockHashesFromNumber (0x08):
[number: P, max-blocks: P]
Version 61 added the BlockHashesFromNumber (0x08) message which could be used to request blocks in ascending order. It also added the latest block number to the Status message.
Version numbers below 60 were used during the Ethereum PoC development phase.
0x00
for PoC-10x01
for PoC-20x07
for PoC-30x09
for PoC-40x17
for PoC-50x1c
for PoC-6