Ok, this is maybe going to be a little hard to follow, but here's an argument why open will come back. A year and a half ago, I drew an ugly picture, intended to describe the hive mind at work. It has been invalidated since then, because we all experience a lot of information digested and made relevant by our network neighborhood - our facebook friends and twitter follows - which we didn't really do two years ago.
The silos have been useful here in establishing a namespace for that neighborhood. The information, however, is poorly aggregated and very ephemeral, not because it has to, but just because that's how the silos are running currently.

The novelty of the namespace will wear off. We will find more open ways of finding our network, and be less dependent on the single provider. And then the next game will be on. How do I keep the information around me? And how does it aggregate beyond my neighborhood. How can I have a soft boundary to a bigger world (i.e. less info, when the information is distant to me) instead of the current hard cutoff?

There's no way the silos are going to be able to squeeze all the value out of the data they hold. There are simply too many kinds of value, and its neither feasible nor cost/risk-reasonable to try to figure out where the value is within one corporate budget.
So the next jump in value of our networked information digestion will come from a reopening - along the lines of the various twitter value-adds that are eating the twitter firehose and condensing that for more value.
If the economics of paying twitter for such an open infrastructure can't be worked out, it'll be made redundant, or paid for by someone like Google who can take out the value elsewhere.

Beneath that argument is the underlying understanding that machines+people provide much more value than machines or people alone; The unattended semantic web still isn't happening - but guided, machine enhanced, local data gathering provides enormous value.
An example is the Roskilde concert-finder I did w. Morten - mixing local info about a music festival with Last.fm's siloed network of music-facts.
The semantic web idea - that we're somehow close to being a world where data of this kind is just magically in the right shape for the machines - is unlikely to come about any time soon.

If we add people to the mix and 4-8 hours of work, we're already there, however. The data just has to be something we can get at.