defaulting to what connect provides currently. This would be easier for middleware such as a connect.timeout() middleware to next(err) after the timeout has occurred, without worrying about where in the stack you might be, as you might have missed all the error-handling middleware. This would be less flexible from a handling stand-point but more flexible for cases like connect.timeout. We would need to flag when an error has already occurred to prevent any subsequent timeouts etc from re-invoking this, however this would also be an issue because these errors would potentially go un-reported which... is not good
gonna close this for now. i think people are pretty used to how error handling is right now and i don't think you'd gain much. plus, i don't want to implement it. haha
if anyone else has suggestions to improve error handling, feel free to speak up.