ARIA (Accessible Rich Internet Applications) is a WAI initiative promoted by the W3C to make rich internet applications accessible to assistive technologies. I'm actually quite a fan of this initiative, but I'm also aware of the fact that maybe we're facing another W3C failure. I'll try to make myself clear. The sad truth about these initiative is that the actual actors of this drama are not on the stage together. The actors of the W3C play are both user-agent vendors and assistive technologies vendors. So far so good.
The problem here is that both actors don't communicate to each other. For example, many of you recall the problem of the CSS declaration display: none interpreted by most screen readers as speak: none. This is a bug, of course, but you can't blame or curse screen readers, because their implementors were not involved with the process behind the CSS specifications. None of them! As result, screen readers didn't implement aural style sheets which were later abandoned. So we have two actors on two different stages, as you can see from the following picture taken from the ARIA specs:
Theoretically speaking, user-agents and assistive technologies should communicate in order to make this model work. Reality is actually different: WAI-ARIA is currently poorly supported by most browsers, not to say ignored by Internet Explorer which is still the most used browser on the web. I think that we're facing a situation similar to the lack of support for application/xhtml+xml which made XHTML 1.1 fall into oblivion (and, consequently, HTML 5 arose from the ashes of XHTML 2). To prevent this from happening, the W3C should put both actors on the same stage by getting them involved with this initiative as much as possible. Otherwise, this would be another W3C failure.