Clean up RenderViewHostManager swapping logic.

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 fcf53c78ccdaf3f846cb5b122653c85730c9bbae..6cfa5ad1a47112b44f3cea9700c06dcd8297c12c 100644
--- a/content/public/browser/content_browser_client.h
+++ b/content/public/browser/content_browser_client.h
@@ -145,10 +145,13 @@ class CONTENT_EXPORT ContentBrowserClient {
   virtual void SiteInstanceDeleting(SiteInstance* site_instance) {}
 
   // Returns true if for the navigation from |current_url| to |new_url|,
-  // processes should be swapped (even if we are in a process model that
-  // doesn't usually swap).
-  virtual bool ShouldSwapProcessesForNavigation(const GURL& current_url,
-                                                const GURL& new_url);
+  // a new SiteInstance and BrowsingInstance should be created (even if we are
+  // in a process model that doesn't usually swap).  This forces a process swap
+  // and severs script connections with existing tabs.
+  virtual bool ShouldSwapBrowsingInstancesForNavigation(

[darin (slow to review), 2012/05/31 04:43:08]

Why pluralize BrowsingInstance here?  Also, it looks like the BrowsingInstance
type is a private implementation detail of src/content/.  Should we really use
that name in the public API?

[Charlie Reis, 2012/05/31 17:58:23]

Seemed consistent with ShouldSwapProcessesForX.  I suppose the other way works
just as well.  Fixed.
I think so.  It's a private class out of convenience (because no one happens to
need it), but it's an important concept.

In this case, it has real implications for the process model: whether or not the
new page has any script connections (even postMessage) to existing pages.  Plus,
BrowsingInstances are described on the process model page and the conference
paper about the process model.

That said, in other APIs we've said things like "GetRelatedSiteInstance" to
imply a SiteInstance in the same BrowsingInstance.  If you felt strongly about
this we could try to find a similar name here, but I think it's going to be more
useful to expose the concept.

+      BrowserContext* browser_context,
+      const GURL& current_url,
+      const GURL& new_url);
 
   // Returns true if the given navigation redirect should cause a renderer
   // process swap.