That makes no damn sense. You're falling for the "example.com == www.example.com" fallacy. It's not true. "google.com" and "www.google.com" are allowed to be completely different servers (content-wise, physically, and administratively). Quoting from that most Holy of SEO Holies, the robots.txt spec [robotstxt.org]:
"local URL", in this (slightly archaic) instance means "relative URI". google.com/robots.txt is not a relative URI for anything on www.google.com, because they've got different absolute roots.
The only robots.txt that applies to a given URI is the robots.txt accessed using the same fully-qualified domain name. That's the spec, and it's the spec for a damn good reason: managing webservers in the .uk TLD (among others) would be chaos if things were done your way.