Brad's solution is to create a service where people go to aggregate all their networks into a master network, and then let other services check against that to automate friend discovery. The outcome to the user who signs up to a new service should be "These 8 friends of yours are already users here, would you like to share your books / music / pictures / trips / etc. with them?"
I'm not so convinced due to the following reasons:
1. Firstly, 'friends' is a nebulous concept and it varies from network to network (my Linkedin friends may not be the same as my Orkut friends), thereby making the whole idea of having a universal social graph quite impractical.
2. Secondly, Brad assumes that all the competing services will actually cooperate with each other to share their respective social graphs. I doubt that will happen. If it did, then it would be equivalent to voluntarily reducing exit barriers for its users.
But it looks like the problem that Brad is addressing is one of singular identity (say a Google Account or a Hotmail Passport) that uniquely maps people across applications. That's a separate problem in itself. Anyway, here is what I think will be the future of social graphs and networks.
Where I see this heading
I forsee that in the future there will be two kinds of 'social network' services:
1. A plain vanilla social network with the usual friend of friend and profile features.
2. Satellite social applications like say a photo sharing service, a book sharing service etc.
My guess is that there would be about 3-4 major plain vanilla networks. The satellite services would then sit on top of these networks and share their social graphs. So say I signup on Facebook (the plain vanilla network), and then I feel like signing up at Shelfari.com to share books with friends. I would just activate Shelfari inside Facebook. The Shelfari-Facebook tieup would not be exclusive. Shelfari could go and signup with all the other vanilla providers too. Facebook has already moved in this direction by allowing 'applications' by independent developers to hook into Facebook. The next level is for all these 'applications' to have their separate identity in the world outside Facebook, thereby allowing users of other vanilla networks to use them too. The following image should help clarify what I'm trying to say.
Thus, the vanilla networks would serve the purpose of being repositories of social graphs that independent developers of services can tap into.
An arrangement like this would be very useful for all web applications with a social dimension to them. The major vanilla networks house the social graphs, and the independent guys hook in as satellites and share revenues with the parent network.