Google provides an iPhone application for Google search. What I found interesting is how they make sure the back button works consistently.

If you’re not familiar with the application, that’s a normal workflow:

  1. Enter a search term.
  2. Along the way you get search results for partial queries and proposed auto-completed queries.
  3. Click any item in the search result.
  4. iPhone’s Safari is now launched with the target page. At this point you leave Google’s application. Because of the application model on the iPhone this also means that the Google application gets closed completely.

Now for the hack that Google implemented. The back-button works in Safari. It does the following actions:

  1. The first time you click the back button you go to a search result page on google.com. This time in Safari.
  2. The second time the Google application gets launched again and the query is executed again in the application.

I was very interested to find out how they do it. And if you watch closely you see that upon launching Safari, Google opens a web site on some IP address. I assume that in that URL they embed a request ID. Google can then count how often the page with that ID has been opened and do different actions based on that:

  1. First time: Redirect the user to the entry he was looking at.
  2. Second time: Redirect the user to the result list.
  3. Third time: Open the Google application. I’m guessing they do this using a registered application protocol.

As they redirect from that temporary page, the back button will always go back to that same URL.

I found that a pretty clever hack. I’m just left wondering why they open this temporary page on an IP address instead of some google.com subpage.