How does Google Friend Connect accomplish cross domain communication without needing to upload a file to the client domain?

Question

Previously, Google's Friend Connect required users to upload a couple of files to their websites to enable cross domain communication and Facebook Connect still requires you to upload a single file to enabled it.

Now, Friend Connect doesn't require any file upload... I was wondering how they were able to accomplish this.

Reference: http://www.techcrunch.com/2009/10/02/easy-does-it-google-friend-connect-one-ups-facebook-connects-install-wizard/

Answer 1

There are multiple methods of communicating between documents on different domains, amongst these HTML5 postMessage, NIX, FIM(hash/fragment), frameElement and by using the window.name property.

These are available on different browsers and in different versions, but collectively they allow you to do reliable XDM (cross domain messaging).

One project that have done this earlier is Apache Shindig, which probably pioneered quite a few of these, and more recently, the project easyXDM has come, unifying all of these approaches with a common API, making it easy to create complex applications using XDM and RPC.

You can read in depth about the various methods of transporting the data in this article at Script Junkie.

Now, to answer your question directly, earlier on it was quite common to believe that there was only postMessage, the FIM (Fragment Identifier Messaging) available, and for the latter to work efficiently, one often had to upload a special file to your domain. As more methods have been discovered, this has by many been deprecated as a technique, and hence; no more need for the file.

Just for the record; I'm the author of both the Script Junkie article, and the easyXDM library (that is what Twitter, Disqus and quite a few more are using by the way).

Answer 2

<edit>It's difficult to remember/verify now, but I believe my answer here was probably incorrect. Sean Kinsey's answer above should be the definitive answer to this question. If you're reading this, please upvote his answer and ignore mine.</edit>

The Google Friend Connect widget works like most ads/gadgets do, using a copy/pasted snippet of HTML to reference a JavaScript include on the host's server which then creates an iframe containing the desired content. By opening the iframe with your site ID in the URL, Google's server is able to generate the appropriate HTML document to represent a Friend Connect gadget for your particular site/settings.

There isn't any cross-site communication happening beyond that initial step of creating an iframe with the appropriate URL target. Everything inside the gadget's dynamically generated iframe is more like the user visited a separate page on Google's server, but what would have been displayed is then embedded/isolated in a block on your page instead.