connections

Documentation

Index

• type Dialer
• type Listener
• type Transport
  • func (t *Transport) Dial(l *net.UDPAddr, r *net.UDPAddr) (UDPConn, error)
  • func (t *Transport) Listen(l *net.UDPAddr) (UDPConn, error)
• type UDPConn

Types

type Dialer

type Dialer func(*net.UDPAddr, *net.UDPAddr) (UDPConn, error)

type Listener

type Listener func(*net.UDPAddr) (UDPConn, error)

type Transport

type Transport struct {
	Dialer   Dialer
	Listener Listener

	//Used to better support backwards compatibility.
	Close func() error
}

Connections handle themselves based on the send from address and the receiving address. This is because handling unicasts and building the connections becomes complicated to handle at the client level as each implementation has different variations. For example packet sockets write to the raw ethernet connection by device id, adding the header information to the packet which is unrelated to the connection. This means when switching from broadcast to unicast it's still the same connection to the ethernet port. Therefore existing connections don't need to be closed between connections to other hosts. This information is only know by the connection type not by the client. Therefore rather than the client handle this the connection package itself is responsible for the implementation details. The idea of having transport here is to allow the implementation to manage pooling much like the HTTP transport in the standard lib.

func (*Transport) Dial

func (t *Transport) Dial(l *net.UDPAddr, r *net.UDPAddr) (UDPConn, error)

func (*Transport) Listen

func (t *Transport) Listen(l *net.UDPAddr) (UDPConn, error)

type UDPConn

type UDPConn interface {
	ReadFrom(b []byte) (int, *net.UDPAddr, error)
	Write(b []byte) (int, error)

	Close() error

	LocalAddr() *net.UDPAddr
	RemoteAddr() *net.UDPAddr

	SetDeadline(t time.Time) error
	SetReadDeadline(t time.Time) error
	SetWriteDeadline(t time.Time) error
}