I feel like I'm taking crazy pills. How on Earth could anyone consider the example in #2 "conflict-free"? You haven't removed the conflict, you're just ignored it! Anything can be conflict free in that case.

Obviously not every problem will have such an obvious right answer, but given the example the author chose, I don't see how you could accept any solution that doesn't produce "100" as a correct result.

▲

WorldMaker 4 days ago | parent | next [-]

I also think this is a place where CRDTs in general got stuck on the name "Conflict-Free" for way too long assuming it was fate that if they worked hard enough they find the magic data structures to eliminate conflicts altogether but real life data is a lot more more complicated than that and real life expectations of data semantics that a data type itself can't encode. I think we are just now getting to the point of seeing CRDT libraries understand some conflicts happen, and some conflicts still need to bubble up to a more complex semantic model or even/especially a user. I don't think there are any CRDT libraries that are strong for that yet, but the work seems to starting into those next steps at least.

▲

dwaltrip 4 days ago | parent [-]

I remember like 10 years ago seeing some people getting really excited by CRDTs. I was deeply confused about how the software would magically know what the correct thing to do would be when 2 people made edits that directly conflicted with each other…

You can’t know what to do without talking to the people involved, as they have to decide what makes sense for end goal. It’s mostly a social / collective action problem, not a purely technical one.

	▲	WorldMaker 4 days ago \| parent [-]
		Yeah, it's hard to fault people from wanting to find as many technical solutions as possible to reduce social problems to a minimum. I can fault CRDTs for thinking that they could solve all of them technically, enough so that they put "conflict-free" in the name, because yeah, there are always problems you can't solve technically, there is always user knowledge you can't just assume the technically best approach is the semantically, socially, or even politically best approach (again as much as we might want to wish that we could solve these things technically, which is a beautiful dream sometimes).

▲

crazygringo 4 days ago | parent | prev | next [-]

Yeah, "conflict-free" is from a theoretical perspective, that it always generates a final result guaranteed to match. Which, to be fair, is an accomplishment in and of itself, when the operations might be out of order or repeated.

It is absolutely not "conflict-free" from the user perspective, nor is it even necessarily logical. Tools like Google Docs manage this by providing a version history so if the merge messes up, you can still go back and see each recent change to grab whatever data got lost.

▲

Aldipower 3 days ago | parent | prev | next [-]

And the example #1 isn't even better. Maybe, I am not smart enough, but when the physical clock of user B is _way_ off in the wrong direction, you still get the wrong ordering. Example 1 assumes that the users clocks are going more or less correct, which is sometimes just not the case.

▲

AlienRobot 4 days ago | parent | prev [-]

Yeah, that's an insane things for a program to do. The only correct thing for a software to do in case of conflict is to warn the user there is a conflict and provide them with the tools to fix the conflict. Assuming a solution is the last thing anyone wants.