It's just as shallow to compare a desktop software application's success to the much more transient, ephemeral, and difficult to quantify success of Facebook. Better to look at the Internet as a whole and ask a couple of simple questions.
First, name one network-wide, user-oriented application level service that was present when the commercial Internet opened for business in 1991 that is still in operation and use today.
Discounting UseNet and Email as infrastructure, the answer is likely "nothing." It's instructive to consider why. Early community plays on the Internet (The Well, The Globe, AOL, WebTV, and even MySpace) fell in succession, not because there weren't plenty of users and not because they weren't good services. They fell because, by definition, something newer and better comes along. It's the same reason we don't drive horse-drawn wagons to work. Supporting the infrastructure and feature set of an existing system means, by definition, that you will never be able to change and adopt new technologies as fast as someone else starting with a clean slate.
Second, what is so special about Facebook that it will avoid being obsolesced by the next cool fad? Answer again, "nothing".
Facebook's only advantage is the depth of its social graph. And as many posters have noted, the average Facebook user has a pretty static social graph and no need to add to it in any significant way now. Once you are fully connected, it becomes trivial to notify your graph that you are moving elsewhere, and then Metcalfe's Law kicks in. Once the infrastructure becomes distributed and you are no longer locked into a single service, people will be free to move their social graph and associated applications wherever they'd like.
Extrapolating the past lifecycles of similar, successful social sites to Facebook, it seems logical to conclude (as the author did) that Facebook's days are numbered. Maybe in the thousands, but numbered nonetheless.