For a long time we’ve been taking the “worse is better” approach to security (ambient authority). This approach has been “working” until fairly recently (perhaps the past decade) for one reason: the number of attackers was small, and therefore security was something you could get away with cutting corners on. Any system is secure in the absence of an attacker.
We are now (especially in the wake of “IoT”) coming to grips with what a terrible mistake cutting corners on security has turned out to be. Capabilities were, like many ideas in computer science, “ahead of their time” and it’s only recently that the problems they solve have tangibly materialized.
In the wake of a mounting cybersecurity threat, I believe we’re now seeing a sort of capability renaissance. Operating systems in high security contexts like seL4 have adopted a purely capability-based model and it’s an approach I can only assume will continue to proliferate in these contexts. Thanks to projects like Capsicum, I hope we will see capabilities leveraged in mainstream operating systems like Linux as well.