As I understand it, the NSA's recently disclosed collection of telephone metadata falls on the right side of that bright line inasmuch as no one has a reasonable expectation of privacy regarding the sort of information that telephone companies routinely collect in the ordinary course of business. That, at any rate, is the holding of the most readily applicable legal precedent, a 1979 Supreme Court case under the caption of Smith v. Maryland. You might quibble about Smith's applicability in the present case on the ground that, owing to modern technology, metadata reveals a lot more about telephone-users than it did in 1979. (See e.g. this Jane Mayer piece). But it's hard to argue that what the NSA is doing is condemned by the law as it now stands.
Why aren't these considerations more reassuring? Here's one answer:
Notice that all determinations about the applicability of Fourth Amendment requirements involve an element of circularity: What a government may do to collect information about its subjects at a given point in time is a function of its subjects' reasonable expectations of privacy. But those expectations, in turn, are a function of what is reasonable in light of what the subjects actually know, or should know, about what their government is doing.
When it's functioning properly, Fourth Amendment law eases the societal trade-off between security and privacy by enabling governments to employ new surveillance technologies in response to new threats while giving subjects a fair chance to adjust their privacy expectations accordingly as they employ alternative modes of interaction. At least in theory, that process of mutual adjustment enables society to trade security off against privacy at the most favorable available rate of exchange.
Seen in this light, searches and seizures become especially unreasonable in two circumstances: first, when the government fakes its subjects out by exploiting reasonable expectations inducing them to act or interact unguardedly in situations in which they're clandestinely being watched; and second, when the scope of the surveillance is so encompassing that it affords subjects too few adequate alternative means of conducting their affairs to enjoy what we regard as an adequate private life.
The NSA data-mining operation raises red flags on both counts. When its advocates claim that the program's secrecy is essential to its effectiveness, they imply that smudging the line between public and private life, and thereby denying subjects the opportunity to adjust their expectations, isn't a bug but one of its features. And the scope of the surveillance, encompassing nearly every available technology of remote communication among subjects deprives them of alternative channels of private interaction. The price we pay in the currency of privacy for whatever increment of security we gain is therefore especially high.