Add time-to-network histogram considering browser side navigation

content/browser/frame_host/navigator.h

Index: content/browser/frame_host/navigator.h
diff --git a/content/browser/frame_host/navigator.h b/content/browser/frame_host/navigator.h
index af8e71ec151a6eb8f33b8e916dc8b3c380096fa3..00864b3199a47b36c190908679257df6dfc3cc59 100644
--- a/content/browser/frame_host/navigator.h
+++ b/content/browser/frame_host/navigator.h
@@ -6,6 +6,7 @@
 #define CONTENT_BROWSER_FRAME_HOST_NAVIGATOR_H_
 
 #include "base/memory/ref_counted.h"
+#include "base/time/time.h"
 #include "content/common/content_export.h"
 #include "content/public/browser/navigation_controller.h"
 #include "ui/base/window_open_disposition.h"
@@ -126,6 +127,15 @@ class CONTENT_EXPORT Navigator : public base::RefCounted<Navigator> {
   virtual void CommitNavigation(RenderFrameHostImpl* render_frame_host,
                                 const NavigationBeforeCommitInfo& info) {};
 
+  // PlzNavigate
+  // Called when the first resource request for a given navigation is executed

[nasko, 2014/09/24 00:48:06]

nit: Is there a second/subsequent resource request?

[carlosk, 2014/09/24 18:35:49]

Can't there possibly be another one? I might be wrong but wouldn't redirects
(and possibly other valid use cases) cause one single navigation request to make
a 2nd network request? Or maybe a "resource request" not exactly what I'm
assuming it is (just a request to the network)?

I'm still too noob in this matter to know it for sure...

[nasko, 2014/09/24 23:11:20]

That is a good point. I was thinking of "navigation" request, which there will
be one, but you consider redirects as separate resource requests. Makes sense.

[carlosk, 2014/09/25 17:26:05]

Cool.

And in fact that makes me think I might need a way to check I'm not handling a
2nd call to this. I talked to clamy@ about it and will check for transfer_ids
(because apparently redirects won't in fact cause this effect). See changes in
ResourceDispatcherHostImpl::OnRequestResource for this.

+  // so that we can time it.
+  // TODO(carlosk): replace the URL as a request-identifier with something more
+  // sure/consistent once we have this something available (PlzNavigate should

[nasko, 2014/09/24 00:48:06]

nit: Let's try to avoid using "we".

[carlosk, 2014/09/24 18:35:49]

In fact, given that this "more precise identifier" will only be available for
PlzNavigate (and not the current navigation) I completely removed this TODO.

+  // provide a request_id).
+  virtual void LogResourceRequestTime(
+    base::TimeTicks timestamp, const GURL& url) {};
+
  protected:
   friend class base::RefCounted<Navigator>;
   virtual ~Navigator() {}