TODO from Nov 3 2013
parent
1aa0420048
commit
54df95eb3f
@ -0,0 +1,29 @@
|
||||
Bug: onmessage:
|
||||
if (j === "ping") {
|
||||
this.sock.send(JSON.stringify("pong"));
|
||||
return;
|
||||
}
|
||||
|
||||
- if the connection closes before the pong can be sent, exception leads to process crash.
|
||||
- thoughts: 1. the websocket should be closed if the process managing it exits
|
||||
2. the websocket-manager process should be supervised
|
||||
|
||||
- how to achieve fate-sharing?
|
||||
- onexit handler for processes? dual to boot?
|
||||
- wait, the socket is being closed correctly. it's the zombie restart that's the problem
|
||||
- make World.wrap() check that the intended process is still alive?
|
||||
|
||||
haha my code is silly
|
||||
it needs to be refactored
|
||||
the connection process should just exit when onclose is detected
|
||||
a supervisor should manage the reconnect exponential backoff
|
||||
and finally, the process that does the connected/disconnected display is looking at the wrong thing
|
||||
it's being too trusting
|
||||
it doesn't tell anyone if the connection itself has vanished
|
||||
only when the connection is present but changing internal state
|
||||
|
||||
For now, I've trapped exceptions sending :( and left the reconnect
|
||||
logic in the WebSocketConnection.
|
||||
|
||||
TODO: make a proper supervisor, with timer-driven hysteresis and
|
||||
conversation-specific state machines.
|
Loading…
Reference in New Issue