As you can guess, netchan makes Golang channel work over network. Both ends of the channel can talk synchronously or asychronously.

Example:

  // a sending process
  ch := sender.NewChannel("my_log")
  for line := range lines {
    ch <- []byte(line)
  }
  
  // a receiving process, on a different machine
  ch := receiver.NewChannel("my_log")
  for line := range ch {
    println(string(line))
  }

Both channels are chan []byte. Of course, sender channel is send-only, and receiver channel is receive-only.

Simple, right? Read on. It's more sophisticated underneath.

netchan has 2 or 3 components: optional Leader, Sender, Receiver

Leader is solely for service discovery. Basically, mapping a name to a service location.

This can be skipped if everything runs intranet, where we just need a simple network broadcast.

Current Leader implementation is a simple http server. Later, service discovery can use Etcd, Zookeeper, etc.

It's actually not Sender sends messages to Receiver, but Receivers pull messages out of Sender.

Sender stores all messages on disk locally for a configurable time, depending on avaiable disk space. This makes asynchronous message passing possible. When a Receiver requests the data by a timestamp, Sender will respond with corresponding data.

If a Receiver catches up, Sender will pass the most recent messages to the Receiver.

One Sender can have multiple Receivers.

Receiver can give Sender a filter, so that Receiver only receives filtered results.

netchan usage is similar to common messaging systems. However, netchan does not need a separate messaging system such as Kafka, NATS, etc. The data is stored on the sender directly.

Welcome any sorts of help, discussion, pull requests, tests, etc.