I don't think I'd characterize his blog post that way. The last sentence captures what I got from reading it: "It may not be as beautiful as federation, but at this point it seems that it will have to do."
Moxie's essay is strong and djb's point is not: a "federated network" with a centrally controlled app that updates automatically might as well not be federated at all: at that point having multiple servers hardly matters. Who cares about running your own server if you can't run your own client?
Moxie has put into words what I've been thinking myself for a long time - the old federated protocols from the 1990s have been slowly dying off (with email being the main holdout, unless you count gmail). And the reason is that they just don't evolve. There's no incentive to evolve them, there's no organisational structure to evolve them, creating them takes huge effort AND they suffer from all sort of hidden ideological constraints that would prevent them from e.g. doing a partnership with Uber beause Uber is a corporation and an open protocol isn't. But users don't care about that.
Also, as djb points out: https://twitter.com/hashbreaker/status/732912508089032706