Further evidence favouring openness (and even the throwaway certificates i mentioned below) can be found in a long and entertaining interview with Bruce Schneier one of the worlds leading cryptography experts.
His contention is that even for that very important function of verifying identity there are no safe measures deployed, and any and all of the grand schemes to do so will fail very often. This tells us two things: First, that schemes that don't have to be grand are better. They too do eventually enjoy a network effect, but they don't require everybody to be plugged in to work. Secondly, interfaces will be compromised, so you better prepare for it somehow, by limiting the consequences per breach.
His point of view is directly related to thought about digital identity and comes out in favor of loose-knit reputation systems and throwaway identification in specific cases to guard against the consequences when (not if) your identification point itself becomes compromised (either because you were careless or for systemic reasons).
However one can't help but feel that even though the computation of identification is the most powerful computation there is, the points should apply to all the other computations also. So in a way I think the Schneier article comes out - indirectly - in favour of the openness of design efforts also. Any idea that we can keep our world closed through the application of technology is flawed, so we might as well build it open from the start. Open with anti-intrusion measures that is.