So far this is probably the best "intro to CRDTs for a developer" I've read. I built a product around CRDTs, essentially, and my god was it painful trying to engage with. Showing actual code, explaining that `merge` is the fundamental operation, etc, is really all a developer needs to know IMO.

Also, the fact that we always use text editing as the de-facto solution is so weird to me since that problem is both niche and extremely complex. IMO a better example would be something like "Can this person drink alcohol?". Age moves in one direction so it has a simple merge function:

    def set_age(self, new_age: int):
        self.age = max(self.age, new_age)

A property of this is that if I query your age and if you're 21 I can cache that age forever. You'll only ever be >= 21, after all. If I add new queries that care about you being 25 (for a hotel) I can satisfy the "drinking age" queries from a stale cache and then retrieve the true value (<25) when I need to check if you can book a hotel.

This means you can have distributed caches without invalidation logic. A pretty amazing property since cache invalidation is a hugely complex problem and has seriously negative performance/ storage implications.

It also means you can drop writes. If my system gets information that a person was 18, but that information is out of date, I can drop that write, and I can do so by examining the cache and viewing stale information, only checking the real value if the cache value is < 18.

This whole thing lets you push computation to the edge, drop expensive writes, ignore any cache invalidation logic, cache values forever, potentially answer queries from stale cache values, etc.

Anyway, kudos for the writeup. I skimmed the second half but the first half was great and the second half looked legit.