Please document how concurrency is handled

[pavlix]

It would be great if it was documented how concurrent requests, especially when Tokio is used, are handled for netlink-proto and also rtnetlink, for that matter. How simultaneous request-response exchanges are handled and dispatched and how event monitoring is done simultaneously with request-response exchanges. It would help to know what we can rely on and what we cannot.

[cathay4t]

You should not use single rtnetlink::new_connection() to host simultaneous request-response exchanges. You can only send out request after reply done or error.

Please create multiple rtnetlink::new_connection() for your use case.

For event monitoring, the rtnetlink/examples/ip_monitor.rs can helps.

[cathay4t]

In theory, rtnetlink crate and kernel code have sequence number to track multiple communication session for a single socket(connection), but I would not suggest so as it might easily overflow the netlink socket buffer especially with BGP routes.