Jakob Nielsen just posted Mobile Site vs Full Site. Like much of his writing, it's provoked a critical response from web designers. Most of this criticism is for suggesting separate mobile sites; suggesting mobile is a separate use case; and not mentioning techniques like Responsive Web Design. I've not read the full report (as, I suspect, most critics haven't), but I think it's a bit unfair.
As he clearly states in the first 8 words of the body of his article ("Based on usability testing of hundreds of sites…"), his advice derives from observations of real user behaviour - often a rich source of uncomfortable truths. NNG observed users accessing existing web and mobile sites on a mobile device, and where there was a mobile-specific version, they found it easier to use. It's rational to conclude from this that there ought to be mobile-specific versions of content.
He is unhelpfully specific as to how this should be done ("If mobile users arrive at your full site's URL, auto-redirect them to your mobile site"). I think if I read that literally I'd be tempted to disagree with the detail of implementation, but find it hard to argue with the intended effect. From the end-users perspective the advice is valid: give mobile users a mobile representation and let them switch representations if there are features on the desktop site not present in the mobile version, that they wish to use.
How this is achieved is a different matter, and a point on which he's quiet, if not silent: leaving room for multi-serving, RWD, RESS, or other approaches (in a follow-up interview by .Net magazine, he suggests all three). A user can't be expected to notice whether the site they're accessing works well by virtue of its being responsive. They are unlikely to care whether it was produced by a server using separate templates for different representations of content, or some other mechanism.
He talks of a "mobile use case", but if you look at the detail of what he suggests it doesn't seem so contentious. To suggest that mobile sites should avoid esoteric features without limiting core functionality is valid: Amazon, Google and eBay all do exactly that; even a textbook example of responsive web design like the Boston Globe hides some content (the "social" links are hidden and revealed by a "More" link next to the Tools section of the footer). If Nielsen is basing his advice on observations of users, I would expect that they have demonstrated such an approach to be a good one.
The challenge, as he clearly points out, is finding the "cut between mobile and full-site features in such a way that the mobile site satisfies almost all the mobile users' needs"; and this is the crucial "it depends" get-out which gives Nielsen and his critics the wiggle room to find agreement.