sachanganesh / connect-rs Goto Github PK

The following code from the TCP server Stream impl and TLS server Stream impl uses block_on within the poll_next function. This is a really bad practice and should be changed.

However, naively using the corresponding poll_x function of the async code leads to misbehavior in the server example programs. The servers just hang and do not get alerted to new connections. This particular error is most likely an implementation error on my part, so the proper solution is unclear.

There could be a use case for creating connection pools to certain destinations that have too high a backlog of messages. Of course, with this added concurrency, there needs to be a way of re-ordering messages on the receiving side.

Improve robustness in the case of an unexpected connection closing by retrying with a certain level of backoff

Before this crate be published and used by others, all source must be thoroughly documented

I'm not sure at the moment how to design reasonable tests for things like the socket successfully bound to a ip-addr/port, etc. without some human verification. This needs to be solved in the long term though.