'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.
tx-type
must be known to the implementation. Definedgas-limit
must cover the 'intrinsic gas' of the transaction.gas-limit * gas-price + value
) of thenonce
of the transaction must be equal or greater than the current nonce of theImplementations 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:
parent-hash
ofdifficulty
, gas-limit
and time
are within the bounds of protocol rulesgas-used
field of a block header must be less than or equal to the gas-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.
header
must be valid.transactions
contained in the block must be valid for inclusion into the chain attx-type
is permitted at the block number isgas-limit
s of all transactions must not exceed the gas-limit
of thetransactions
of the block must be verified against the txs-root
by computing andommers
list may contain at most two headers.keccak256(ommers)
must match the ommers-hash
of the block header.ommers
list must be valid headers. Their block numberReceipts 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.
Version 66 added the request-id
element in messages GetBlockHeaders, BlockHeaders,
GetBlockBodies, BlockBodies, GetPooledTransactions, PooledTransactions,
GetNodeData, NodeData, GetReceipts, Receipts.
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.
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.
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:
[hash: B_32, max-blocks: P]
[hash₁: B_32, hash₂: B_32, ...]
[hash₁: B_32, hash₂: B_32, ...]
[[header, transactions, ommers], ...]
[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