XHTML won't go away, you will still be able to use XHTML 1.0 Strict (or XML of course) to format and structure your information in the purest way and use it in combination with CSS to create layouts for various media, it will work in the future as it works today.
I never said that it wouldn't.
When it comes to document creation and structuring information, HTML 5 brings some content-specific additions, like ,
And that stuff is pretty reasonable, if maybe a little over-specific.
But beyond that, HTML 5 brings lots of improvements for exactly what you moan about here – application development! Local data storage, Local file access, Local SQL database, Application cache (much enhanced speed and offline usage), Javascript workers, and of course XHTMLHttpRequest 2 is all brought to the table with application development in mind, and in combination with the canvas, inline SVG, WebGL, the new media tags and CSS3 it will mean a great revitalization of the "AJAX" concept and there is no doubt you will see a "version 2.0" of "Web 2.0" in a near future! 
Yes, I
know that HTML5 adds features intended for application development. My point wasn't that it didn't have features to support it, my point was that
adding features to support it is a fundamentally misguided and stupid idea because HTML was never intended to be an application platform in the
first place. It's like putting weapons hard-points on a tricycle - if you're doing it, you obviously don't understand what a tricycle is for, and you're just going to bog the thing down for its intended purposes.
HTML is never going to be a good application platform because it was never intended to be an application platform at all, and because you can pick any two browsers (or many combinations of
versions of the same browser, even) and your site will not look or work the same across both. And adding features intended for application development is going to make it much easier to create terrible, terrible webpages. Handing over things like local file access to J. Random Web Developer is
madness from a security standpoint (how long before we have ad scripts that deposit links to client websites on your desktop? We're certainly going to need canvas-blocker plugins in the near future.)
Oh, and what does "Web 2.0" mean
this month? Has it stopped being a vague, meaningless buzzword yet and started referring to any specific thing?
I think you vastly underestimate the importance of the possibilities with HTML5/JS/CSS3 as well as the demand for "Web 2.0" kind of applications. The possibilities are enormous and is just about to get better by a magnitude.
I think you vastly
over-estimate it. There is essentially nothing that can be done with HTML5 that wasn't already done with Flash, and people stopped wanting Flash applications sometime around ten years ago, when they realized that browser-based applications are obnoxious, slow, and practically never fill a need that there isn't perfectly decent native software for. That should be a lesson, but the people behind HTML5 aren't interested in lessons; learning lessons would require acknowledging that someone might know more than them about something.
"The possibilities are enormous?" Well, that's arguably true - but it's never been about
possibilities, it's been about
results. You
can use an application made with HTML5/JS, but it's only going to be as fast as your browser allows, and only going to have as coherent a UI as the quality of your browser's rendering engine and the constraints of building a graphical user interface out of webpage elements will allow. Alternatively, you could just find native software that does what you want, which will be far better able to harness the power of your system, and which may even have a GUI designed by someone who knows about more than how to finagle CSS into rendering consistently across a few of the most mainstream browsers. Gee, I wonder which is the better option.
I will follow these kind of developments with great interest, although I think it's a bit misleading to speak of "Web OS" when all there is to it is what the user look and feel, the UI, etc (and not the underlying OS). But OTOH, this is the only important thing to 99.9% of the users. As a user you never see any low level stuff. You can't see a driver, you can't see a file system. The users may experience the *results* from stuff like that being present, but they will do it through the UI, which I guess will be handled by HTML5/CSS3 empowered AJAX.
I can't see the wind, either, but I can sure as hell feel its effects. Javascript is a crappy language for application-scale development, replete with design choices that encourage bugs, bloat, and plain old bad coding. It's also an interpreted language, which is and will always be slower than native code. HTML5 is loaded down with features that should never have been introduced to the standard, heavier than its predecessor (which was no lightweight to begin with.) The combined overhead of all this is going to have a noticeable effect on performance, especially on mobile hardware, and even an ordinary user can sure as hell tell when the computer is being slow.
And an ordinary user can definitely tell the difference between a UI created with a proper UI toolkit and a UI cobbled together out of HTML and CSS elements, if nothing else.
Many possibilities with this IMHO, and I have seen many demos and examples that has convinced me that not only is it possible, but it may actually be quite good.
Again, possibilities are meaningless. Given sufficient memory and horsepower,
anything is possible in
any Turing-complete language. You could write full-fledged applications in INTERCAL, too, but that wouldn't make it a good idea. And even assuming that the results
were good (which would require
all of the following: an efficient Javascript JIT, an efficient, strongly standards-compliant page rendering engine, and a developer who can create solid, efficient, low-bug software in a language prone to none of the above - and only the last factor is even something that the developer can control,) it would
still never be any better than what native software can achieve.
(BTW, there are some links in this first post you can take a look at! 
Yes, there certainly are. And
every single one of them would've been done better with either native software or ordinary HTML4. There's a lesson in there, somewhere.
