Add RAPPOR metrics of ServiceWorker.ControlledPage

content/public/browser/content_browser_client.h

Index: content/public/browser/content_browser_client.h
diff --git a/content/public/browser/content_browser_client.h b/content/public/browser/content_browser_client.h
index fbe8214fd2462901e9ebf31f5630974233bfb58a..f0c7eaf32924a70429050e77a2a2ad35af3f1aff 100644
--- a/content/public/browser/content_browser_client.h
+++ b/content/public/browser/content_browser_client.h
@@ -626,6 +626,10 @@ class CONTENT_EXPORT ContentBrowserClient {
                        const OpenURLParams& params,
                        const base::Callback<void(WebContents*)>& callback);
 
+  // Records the domain and registry of a url to a Rappor metric.
+  virtual void SampleDomainAndRegistryFromGURL(const std::string& metric,

[falken, 2015/02/25 01:43:59]

What's the reason we're recording domain and registry vs origin?

[horo, 2015/02/25 02:27:00]

Rappor is designed to use "the eTLD+1 of the user's homepage".
http://www.chromium.org/developers/design-documents/rappor

[falken, 2015/02/25 03:36:57]

I think that's just saying "Settings.HomePage2" uses eTLD+1 but it's not a
design restriction of Rappor itself. Anyway, domain/registry seems sufficient. I
was just curious because SW are per-origin. Maybe Steven Holte can chime in.

[Steven Holte, 2015/02/25 18:30:12]

It's possible to record things other that eTLD+1, but I would generally
recommend that approach.  In order to detect a result, we need to see the same
value reported from a significant proportion of sources.  The more you subdivide
the smaller the percentage of what your report will actually be estimable.

+                                               const GURL& gurl) {}
+
 #if defined(OS_POSIX) && !defined(OS_MACOSX)
   // Populates |mappings| with all files that need to be mapped before launching
   // a child process.