The P2P Version Message: Byte Layout and Payload Specifications

The version message is the foundational payload that initiating peers transmit upon opening a TCP connection. It acts as the negotiation contract, detailing the node’s protocol capabilities, system time, network addresses, user-agent details, and block synchronization heights.

This guide details the complete byte serialization layout and field specifications of the standard version message payload.

---

📐 1. The Serialization Blueprint

The version payload contains 11 distinct fields, varying in size. Below is the precise sequential schema:

Offset (Bytes)	Field Name	Data Type	Endianness	Description
0 - 3	Version	int32_t	Little Endian	Identifies the protocol version (e.g., 70016).
4 - 11	Services	uint64_t	Little Endian	Bitfield of service capabilities (e.g., NODE_NETWORK = 1).
12 - 19	Timestamp	int64_t	Little Endian	Epoch system timestamp in seconds of the sending node.
20 - 45	Addr Recv	net_addr	Big Endian	Network address details of the receiving node (peer).
46 - 71	Addr From	net_addr	Big Endian	Network address details of the sending node (self).
72 - 79	Nonce	uint64_t	Little Endian	Cryptographic random number to identify self-connections.
80 - ?	User Agent	var_str	Raw (ASCII)	The name of the client software (e.g., /Satoshi:25.0.0/).
? - ?	Start Height	int32_t	Little Endian	Current active tip block height of the sending node.
?	Relay	bool	Raw	Flag (0 or 1) indicating if node should receive unconfirmed txs.

---

📡 2. Detailed Field Analysis

A. Network Address Serialization (net_addr)

To support unified IPv4 and IPv6 networks inside a standard schema, net addresses are always serialized as 26-byte structs: * Services: 8 Bytes (uint64, little-endian) outlining recipient capabilities. * IP Address: 16 Bytes (big-endian). * Port: 2 Bytes (uint16, big-endian network byte order).

IPv4-to-IPv6 Mapping Format

IPv4 addresses are padded inside the 16-byte IP field using the standard mapping syntax: * First 10 bytes: set to \x00 * Next 2 bytes: set to \xFF\xFF * Last 4 bytes: the actual IPv4 address.

$$\text{IPv4 Mapping: } \underbrace{\texttt{00 00 ... 00}}{10\text{ bytes}} \underbrace{\texttt{FF FF}}{2\text{ bytes}} \underbrace{\text{IP Address}}_{4\text{ bytes}}$$

For example, 127.0.0.1 is serialized as: $$\texttt{00 00 00 00 00 00 00 00 00 00 FF FF 7F 00 00 01}$$

B. User Agent Variable String (var_str)

The User Agent is stored as a variable-length string. It starts with a CompactSize integer (1–9 bytes) defining the length of the string, followed by the ASCII characters.

For example, the string /Satoshi:25.0.0/ is 16 bytes: * Length: 0x10 (16) * Value (Hex): 2f 53 61 74 6f 73 68 69 3a 32 35 2e 30 2e 30 2f

C. Self-Connection Detection via Nonce

When a node broadcasts its IP, it might accidentally initiate a connection to itself. To prevent infinite loops: * The node generates a cryptographically secure random uint64_t Nonce during startup. * This Nonce is included in every outgoing version packet. * If an incoming version packet contains the node’s own active Nonce, it detects the self-connection and closes the socket immediately.