The problem with that logic is that Stallman missed a huge point. If, from his example you're using Google Docs, even if the JavaScript is "freed" using his new standard with stylized comments and the @source directive - you are still accessing non-free server software (the Google web servers) that responds to the AJAX requests. Not only that, but your browser is also making a call to the Google Ad server, which also has non-free software. You might also argue that its being served by a modified version of MySQL thats non-free, and perhaps even the firewall and the proxy that its passing through is a custom version written by Google Engineers (likely.)
There are two problems I can perceive with your argument, though:
1. It is still potentially very useful to you to be able to modify the software that runs on your computer, and to share these modifications with other people. This is one of the major points of the GPL.
2. You're describing here a system with three kinds of compoments: (a) client software, (b) server software, (c) server data. It's much harder to argue that (b) should be free software, especially if it's in-house Google software that we're talking about, not distributed outside the company. And (c) is not software at all, so the argument doesn't apply. Should the GPL have clauses that forbid, say, a GPL-licensed web browser from being able to connect to a web server running a non-free http server? What if it's a free http server connected to a non-free database? What if the http server and database are free software, but the people who operate the server don't allow you to download all of their data in bulk and serve it yourself?
You have to draw a line somewhere here, and drawing the line between (a) and (b) seems reasonable.