Add time-to-network histogram considering browser side navigation

content/browser/frame_host/navigator_impl.h

Index: content/browser/frame_host/navigator_impl.h
diff --git a/content/browser/frame_host/navigator_impl.h b/content/browser/frame_host/navigator_impl.h
index 2d87a5b7c3b5a63b2ba0d12733149f6cbdea7429..fb5dbf82ff31aa2397ffc34937171ed96d5a2451 100644
--- a/content/browser/frame_host/navigator_impl.h
+++ b/content/browser/frame_host/navigator_impl.h
@@ -6,9 +6,11 @@
 #define CONTENT_BROWSER_FRAME_HOST_NAVIGATOR_IMPL_H_
 
 #include "base/memory/ref_counted.h"
+#include "base/time/time.h"
 #include "content/browser/frame_host/navigation_controller_impl.h"
 #include "content/browser/frame_host/navigator.h"
 #include "content/common/content_export.h"
+#include "url/gurl.h"
 
 struct FrameMsg_Navigate_Params;
 
@@ -78,6 +80,9 @@ class CONTENT_EXPORT NavigatorImpl : public Navigator {
       RenderFrameHostImpl* render_frame_host,
       const NavigationBeforeCommitInfo& info) OVERRIDE;
 
+  virtual void LogResourceRequestTime(
+      base::TimeTicks timestamp, const GURL& url) OVERRIDE;
+
  private:
   virtual ~NavigatorImpl() {}
 
@@ -104,6 +109,13 @@ class CONTENT_EXPORT NavigatorImpl : public Navigator {
   // events. Can be NULL in tests.
   NavigatorDelegate* delegate_;
 
+  // The time the latest navigation request started, used for a few histograms
+  // under the Navigation group.
+  base::TimeTicks navigation_start_time_;
+  // The URL of the navigation request to allow the later matching with the

[nasko, 2014/09/24 00:48:07]

nit: leave an empty line above the comment

[carlosk, 2014/09/24 18:35:50]

Done.

+  // active when recording histogram data.

[clamy, 2014/09/23 21:54:07]

s/the later matching with the active/later matching

[carlosk, 2014/09/24 18:35:50]

Done.

+  GURL navigation_start_url_;

[nasko, 2014/09/24 00:48:07]

Why not use a pair of (url, frame_tree_node_id)? That will disambiguate much
better than just the URL?

[carlosk, 2014/09/24 18:35:50]

I guess not because:

* I don't think it'll help in the case of navigating twice to the same URL as
I'm assuming the frame_tree_node_id won't change between them.

* And as we only care about main frame navigation, if one navigates to the same
URL in two different main frames they will have different NavigatorImpl and
RenderFrameHostManager instances anyway, right? So the frame_tree_node_id
wouldn't help either.

Am I missing something?

[nasko, 2014/09/24 23:11:21]

I wrote this comment before asking about main frames. Since we only care with
this about main frame timing, we will definitely have unique Navigators between
tabs.
What happens though if we start two navigations to the same URL quickly one
after another? I think it is fine, right?

[carlosk, 2014/09/25 17:26:06]

This is exactly the point that davidben@ brought up and the reason we're not
completely happy with using the URL as the identifier. And yes that situation
can  happen but we're hoping it to be rare enough to not change significantly
the results.

From my understanding, for PlzNavigate we will have a consistent "request id"
available throughout the navigation at some point in the future. But for the
current navigation implementation this is the best we can have (without adding
too much extra cost to this task).

+
   DISALLOW_COPY_AND_ASSIGN(NavigatorImpl);
 };