I created a similar open source javascript/ajax/xhtml only browser-based desktop, with an OS X like GUI. It was fun, but a waste of time, and it never caught on. There isn't much of a use case for extendable OS like browser desktops imho.
Can I suggest changing the font size and/or background? It was extremely difficult to read. Awesome project however - ran really smoothly here in Chrome on OSX.
Your major problem is (anti)anti-framebusting. I'm not sure that can ever be fixed. The solution is for a local desktop manager to have the same programability as a browser but separate the top level browser DOM from the local desktop DOM. Many of these FreeDesktop specs would already be usable: http://www.freedesktop.org/wiki/Specifications
I would suggest that when making a window "fullscreen" to make it position (absolute|fixed) with top, left, bottom, right at zero. You'll get "Free" resizing when the browser changes size.
Quite a few years ago I remember there was a browser Desktop based on Javascript. Critically, it was Open Source and initially quite a lot of people logged on and got involved. The main reason so many people were initially interested is because they opened and documented their API and had facilities to really easily generate applications (even a preschooler could do something). It was possible to clone an existing app for modification just by clicking a button. The source code editor was right there in the desktop too.
At first people wrote oodles of minimal apps which did totally trivial things. Many clones of the same "Hello, World" app. Then the whole thing got overrun with people putting existing Javascript games hosted elsewhere into IFrames.
Some of the problems included the fact that at the time Javascript was too slow. They also weren't able to develop the site fast enough to keep up with what people wanted to do with it. They were spending a lot of time trying to optimise client server interaction. Performance became a real issue.
The most interesting aspect of such a project is the ability to interact with other users of the desktop. But this needs to be at more than just the usual social media level. People already have social media apps in their browser and on their ordinary desktops. Putting a chat client in is not enough.
Developers need to be able to trivially add social capabilities to their apps and it needs to be possible for multiple users to work simultaneously with the same data in an interactive fashion, e.g. work collaboratively on code, drawings, documents, etc., in real time. But it can't just be a free for all. People need to be able to set up fixed groups of individuals who have access to the given data streams.
It would also be interesting to be able to easily set up scientific applications which can do distributed computing by farming out jobs to users of the system who want to contribute cycles to science.
Multiuser games also spring to mind. Basically anything where there is real time interactive content. These need not be complex to be attractive. A game where the only way to progress is to share information with other users, say via an in-game usenet system, would be just as addictive as some kind of multiuser RPG.
The one thing people go to the web for is to meet other people. Any successful Web desktop must make every feature of the Web desktop essential and collaborative. If the feature is personal and not something you want other users to be able to fiddle with then it may as well remain on one's personal desktop and does not belong "out there".
The difficulties in making such an experience seamless and smooth are substantial. Every Web Desktop out there has totally missed the point. They are closed source, so no casual developers are interested, and there's no money in it yet, so no serious developers are interested. They don't provide compelling benefits over traditional web pages and desktops and there's nothing addictive about them that gets you going back to continue your interaction with the system. They just end up being slow, boring, insecure, broken, work tools.
....Ha! That's pretty funny. I just discovered that the project I referred to above was funded by Y-combinator {turns red}.
It was called YouOS. I think the company was WebShaka, which according to their web page either was or is funded by YC. http://www.youos.com/
I liked this project quite a bit.
Actually, I'm not sure if it was fully Open Source or not. But it certainly had an open api and had plenty of little apps whose source code you could view.
It looks nice, but the major problem is of course is people seem to not really want this. Web desktops were built over 10 years ago and never caught on.
And with mobile devices these days, the desktop computing model (overlapping windows, menus etc) is being replaced with modal applications.
Its definitely very appealing. If he could sync with Dropbox or iCloud, that would really benefit people on the move.
My take on "this-was-done-didn't-work" is that those systems were built on flash, and extremely clunky and slow. One of them was even running Windows virtual machines, and throwing screen grabs down to its clients. There is always room for a good useful product (which has had a not so glamorous past).
In 2000 I worked for WebOS.com (site still exists, but you wont find a working product), developing a pure JS desktop environment much like the linked page. The development API was loosely based on Swing and sported a full SDK. There was no Flash involved at all, it was entirely DOM based and employed numerous techniques to improve speed. We even had a system for doing JSONp data communication, back before anyone even know what JSONp was.
Once the thing loaded it was wicked fast, as fast as any native interface, and this was back when IE6 was the top dog and JS engines were horribly unoptimized. If the code still worked it would probably scream in Chrome.
Technological limitations were not the problem, usefulness was. We completed the product and discovered that nobody had any idea what to do with it or a way to monetize it. We considered selling it as a foundation for other companies to build systems on (medical would have loved it), but we had no buyers. Then the dot.com crash happened and it all just went away.
People do want this. You want this. It's a desktop and you use it every day. It doesn't do what you or your friends want exactly and is bloated ... you should be able to collaborate on the design (toolbars, icons, menus, trashcan, background, taskbars, etc) using universal stuff like HTML5 + Javascript (feel free to debate the "programability" of those). It can run locally just like your current system does ... it just needs to be completely separate from your browser. It's your brain, man.
Cloudo (see beta.cloudo.com) has been providing a huge WebOS environment for a years. A very advanced API, dozens of apps including e-mail, rss reader, games... And a very well designed robust API based on XSLT.
But it's also quite inactive for 2 years. Nobody is interested at running current desktop on cloud. Because current desktop environments suck. They're broken.
http://demo.sierra-os.org u: demo p: sierraos (takes a few secs to load)