Open lgrahl opened 6 years ago
May I ask what's the current state on this? What's currently happening in such a case you mentioned in this comment?
On the Websocket layer, the only solution is to send back a "acknowledge" message. Since SaltyRTC is based on websockets, it's up to the applications to ack messages. So, removing send-error
makes sense, I would note in the docs though that websockets don't work like TCP in this case.
Like you said, I do not see much value to add a new field to the disconnected
message. In the end, it does not matter and the client cannot really conclude if a message was received or not.
It's currently implemented but implementations should not rely on it. We still need to remove it.
It's currently implemented but implementations should not rely on it. We still need to remove it.
You mean applications who use SaltyRTC should not rely on send-error
, but implement their own ack protocol if needed?
Yes
The
send-error
message is not really helpful since it cannot reliably tell whether a message has been relayed. See this comment for a detailed explanation.I propose to remove the
send-error
message. We could add a new field to thedisconnected
message that indicates whether the client disconnected (or has been disconnected by the server) gracefully with a close code or the connection has been lost (either with a ping timeout or a TCP timeout/error). But even that may be prone to error and the value is questionable.