The trouble comes when you segment content to subdomain.site.com. The domain authority engines assign to a root domain may not always pass to all the subdomains, so you end up with subpar rankings to what you could earn.
As an example, we launched the Beginner's Guide to SEO on guides.seomoz.org and moved it after a few months to www.seomoz.org/beginners-guide-to-seo (where it currently lives). It had earned tons of links prior to that move, but the day after the 301 back to the main domain (away from the guides subdomain), it jumped to #1 rankings for all sorts of terms, including "seo guide" (recently got beat out by Google's SEO guide and is now #2). This was a good illustration of the power of keeping content in a single sub/root domain location for us.
I have been meaning to blog this - this leads to cloaking exploits. For eg. Google will figure out the shorter canonical link for a domain. So if you have www.lakers.com redirect to lakers.nba.com Google will treat 'lakers.com' as the canonical link and rewrite all the search results.
This can be exploited by registering a shorter domain and then pointing it to the real site, then waiting for Google to redirect it. For eg. Fred Wilson noticed that the guy who owns avc.com was redirecting to his blog. He thought the guy was being a good samaritan, until I pointed out that it was a cloak attack and he was trying to jack all of avc.blogs.com's Google juice.
You can avoid this by not letting Google or Bing setting the canonical and by specifying it with a rel='canonical' link or using Webmaster Tools. Do you know anything else about this Rand or have you seen it in the wild?
Good that this doesn't work
some URL parsers have caught up to non-www domains if you have a .com domain, but try it with an .io domain. it fails most of the time.
that means there are some interesting hipness tradeoffs
* if you have a cool .io domain
* you should have an uncool www
That said, I don't want to suggest this is always true all the time or that Google/Bing never group subdomain authorities together, just that we've seen inconsistency and thus I'd recommend keeping them together.
On the other hand, I think I disagree with the buying a new domain to target an exact match. That will just give signal to Google that it's a new domain and it was unnaturally optimized.
Today this does not work if there is any competition at all for a word. (I've still seen it work for long tail terms with no competition).
Long story short - you need good content to rank well :-)
A user searching for exactly what your website is offering, using the words that make up your website url, should at the very least have a strong opportunity to arrive at your website. If Google did anything else it would offer an opportunity for Bing et al.
Obviously, if you set it to be just example.com, don't have all your promotional materials link to the other, and somewhat less obviously, please be aware that what you set as preferred for Google doesn't necessarily mean squat to Yahoo, Bing, AltaVista or whomever.
I'm not as qualified to touch on the rest, but thought this might be at least a little useful.
So, about that whole "subdomain" thing, as mentioned in this article. The problem in SEO is that separate subdomains accrue separate Google scores, correct? So if I was to serve my whole site from a third-tier domain, like "www.example.com", that would be okay? Provided I were not foolish enough to also put stuff on "example.com" and "other.example.com" and expect links to that content to contribute to the reputation of "www.example.com"?
Basically I'm looking for confirmation that the classic old "www" prefix is okay, if used carefully. (Lots of people hate it for aesthetic reasons, and I used to agree with them, and then I had to dig into the rules for DNS CNAMEs, and the terrifying results have made me fall in love with "www" again. Unless you tell me it's bad, in which case I guess I'll just have to take to drink.)