Add a InitializeNetworkSessionParams method IOThread to remove duplicate code.

chrome/browser/io_thread.h

Index: chrome/browser/io_thread.h
diff --git a/chrome/browser/io_thread.h b/chrome/browser/io_thread.h
index de61126a3955a407e8353be7a420a82da74d5f5a..13530465a9f6ae6a970f228c60d1fa4afb5ccb7f 100644
--- a/chrome/browser/io_thread.h
+++ b/chrome/browser/io_thread.h
@@ -17,6 +17,7 @@
 #include "content/public/browser/browser_thread.h"
 #include "content/public/browser/browser_thread_delegate.h"
 #include "net/base/network_change_notifier.h"
+#include "net/http/http_network_session.h"
 
 class ChromeNetLog;
 class CommandLine;
@@ -82,6 +83,9 @@ class IOThread : public content::BrowserThreadDelegate {
     Globals();
     ~Globals();
 
+    void InitializeNetworkSessionParams(
+        net::HttpNetworkSession::Params* params) const;

[Ryan Sleevi, 2012/11/30 00:10:05]

Globals is already a public object, so is there a reason this interface needs to
leak into the header, as opposed to being an internal method in an unnamed
namespace?

To put it differently: Is there anyone, beyond this .cc file, who would expect
to use this?

[Ryan Hamilton, 2012/11/30 00:17:59]

I would only expect the .cc file to call this method.  That being said, if I
don't make this a method of Globals, and instead made it a method in an unnamed
namespace, it would require sticking "globals." onto each one of the
assignments.  *yuck*  I much prefer the shorter lines.  But if you feel
strongly, I'll move it.

+
     // The "system" NetworkDelegate, used for Profile-agnostic network events.
     scoped_ptr<net::NetworkDelegate> system_network_delegate;
     scoped_ptr<net::HostResolver> host_resolver;