If a peer stops writing the body of a request or response, a Hyper client or server has the ability to detect that and handle it however it likes. However, the same is not true if the peer stops reading the body of the request or response that Hyper is sending to it - the write will just indefinitely hang waiting on space in the stream.

There should be some way for consumers to tell Hyper to abort the write. For example, there could be a poll_abort method on Body which Hyper will poll even when it can't request another body chunk. The body implementation could maintain whatever timeout logic it wants internally and ensure the task is woken up at the appropriate time.