The Conflict Free Replicated Sets

A grow only set is able to be modeled as just a set. And then it's very straightforward to compute the merge function for that. But when we try to have a set where we can remove things, the set becomes more difficult and we run into the same problem: how do I model removals? So there's something called a two P set, a two phase set. It uses two sets to track additions and removals. More of places an interesting invariant on use of the data structure that says one something has been added and removed, it can never be re added.