Regardless of what I write below, the page is nice and works, apart from a few small quirks. And yes, we should have it on sourceforge, too. Or if any of our admins ever get time to do it, redirect the main domain to sf.net.
Ok, are we offtopic or not... probably yes. But as it all is related to the page, maybe not so fatally?
I am not saying the validator is wrong. I am saying it went too far. 128 warnings, and how about showing me the errors? I have to search in the output? Why could the "official" validator do that and this one not?
This is critcism of the program, not what it says about the page. All the errors are of course there. But the real number is 7, not 128.
On the subject of xhtml. Five years ago I would recommend it and argue. Three years ago I would suggest it as one of two alternatives, both with their pros and cons. Today I am not sure anymore what to think

I like how clean its rules are. But you get definitely less problems with strict html. The need to serve it with different mimetype (in http!) is the biggest obstacle. Switching to xhtml is not just "upgrade", it is also a slight shift of paradigm, and if people don't adjust to that, their xhtml will be always $^&%, because they don't think in terms of it. And that is the hard part.
Once it is written as xhtml, let's not break it just for the sake of it.
Should I try and tear apart the sources as well?
html > body > div #green > div #main > div #header > div .center > div .header_nav_menu
Why is the link bar five divs inside each other? Or rather, is there a reason to make it so? And why is the whole page in one div? You have the same background rule for div#green and body. And zero use of cascading, so it is all just "styles". That's why you need so many divs.
Anyway, that is all under the hood, and we don't have to care about that, as long as it validates, which is just a few clicks away.