I understand the rule that, if client and server both support persistent connection, they can use it through a Connection:keep-alive header in the first request. After that, both client and server will still keep the underlying TCP connection open when they are done with the first request/response, and then use the same connection in the following requests/responses.
What I'm not clear about is the programming model. Consider the following client code in go:
resp, _ := client.Get("http://www.stackoverflow.com")
// do some other things
resp, _ = client.Get("http://www.stackoverflow.com/questions")
As far as I know, keep-alive is the default strategy in HTTP/1.1.
Q1: Do these two requests use the same TCP connection?
On the server side:
When one request comes, the Go HTTP framework dispatches it to a handler, then due to keep-alive, the framework should prepare for the next request on the same TCP connection. However I don't see any block-mode read code in a handler. So,
Q2: Does the Go HTTP framework use some kind of non-block mode to deal with keep-alive?
I mean, a handler will not block for reading, but just return when done with a request, then the framework will poll each non-block TCP connection, if one of them has data arrive, it dispatches it to an associated handler, and so on. Only when the request has a header of Connection:Close, the framework will close the TCP connection when the handler returns.
