I had been looking at the Web 2.0 world as a series of data silos, web services, and aggregators. I think that there is another way to look at is as well which is to examine which services stretch across all sites and gain network effect by doing so.
I think that a reputation system is a great example of a service that does this and that gets stronger the more sites it is deployed across. Brad Burnham from Union Square Ventures has a post about why it won't work. Mary Hodder was at a Union Square Ventures session in NY where they talked about it:
Mary - "you can pull data for reputation from Ebay.... but the thing about the difference between what Tim was talking about, maps, and Ebay's reputation information is that the mapping data makes sense when you pull it out of the system, whereas the reputation data, because Ebay is so skewed, it's such a bizarre social environment, everybody is under tremendous pressure to make this sort of, you know, A+++ best sale I've ever had, which, I mean, would only exist if the guy who was selling you the thing drove me the item from Kansas or something, otherwise it's just probably B+. [So]... the reputation information is perfect or it's terrible, and when you pull it out of that system ... it doesn't match up. It doesn't translate with other walled garden reputations."
I have to say that I disagree with Mary Hodder on this one. I think that it would be tremendously valuable to know what someone's eBay rating was in other contexts.
Let's say I'm on Craigslist - right now I have no trust mechansim. I'd much rather at least have the eBay one even though it might be imperfect. I also think that people will understand that when a rating is presented as an eBay rating that they will be able to contextualize what that really translates into.
The more the reputation system is used the stronger it gets. A reputation system is going to be very important in further enabling person to person transactions.