Never, ever, never, ever use timestamps for conflict resolution.
Conversation
I've had this argument so many times and decided timestamp at the time the change was actually made is good enough. Let's talk!
2
1
1
For sure if you can't share a clock within an acceptable delta of accuracy then yes it's a non starter
3
I can see how “syncing a user’s data across their two machines” is fundamentally different from a 10M user DB, yeah.
1
maybe, maybe not. Do your conflict resolution rules involve deleting user data? Does time affect that?
1
Replying to
Yeah, I vastly prefer “parent hash” regimes as in operational transforms.
Replying to
I’m definitely pro-causality. Time is an illusion, lunchtime doubly so, etc, etc.
1
1
1
Show replies