added httptrace redirects handling

[jronak]

#23

[rakyll]

I was thinking whether we should consider each redirect as a new request and report them independently.

[jronak]

Wouldn't that become very confusing, though we can still go ahead with it and display the breakdown for each request. What do you suggest?

[rakyll]

Don't we need a custom RoundTripper to properly measure this?

type traceTransport struct{
    current *http.Request
}

func (t *traceTransport) Current() *http.Request {
    return t.current
}

func (rt *traceTransport) RoundTrip(req *http.Request) (*http.Response, error) {
        t.current = req
        return http.DefaultTransport.RoundTrip(req)
}

latencies := make(map[*http.Request]struct{
     dnsStart, connStart, resStart, reqStart, delayStart time.Time
     dnsDuration, connDuration, resDuration, reqDuration, delayDuration time.Duration
})

tr := &traceTransport{}
client := &http.Client{Transport: tr}

traceClient := &httptrace.ClientTrace{
            GotConn: func(info httptrace.GotConnInfo) {
                        // use tr.Current() to modify latencies.
            }
}

Is it an overkill?

[jronak]

I think if we skip the dns completely as you mentioned in #12 then this would be an overkill as we would have already got the final endpoints link. Else round trip implementation you mentioned above sounds good

[rakyll]

I am in favor of #12 given the core goal of the tool is not to report DNS.

[jronak]

How should we go ahead.
Resolve it by following the redirects?
Resolve the IP in the requestor before making requests?