Thanks, great read! Once upon a time I stumbled across https://tools.ietf.org/html/rfc677, which introduces a concept of eventual consistency, or convergence, back in 1975. They didn't call it that, but the "Consistency" section sounds a lot like it. According the Lamport, this is the paper that inspired him to write "Time, Clocks, and the Ordering of Events in a Distributed System" because he found flaws in the RFC, which may be my favorite example of https://xkcd.com/386/.
Thanks for the great write-up, Pat! https://jepsen.io/consistency (inspired from Peter Bailis's HAT paper) offers at least one answer to your question on when causal consistency can be an amazing gift, and it's when the system designer is building a "Sticky Available" (available under partition if clients only talk to the same server instead of failing over to another server) system.
Thanks, great read! Once upon a time I stumbled across https://tools.ietf.org/html/rfc677, which introduces a concept of eventual consistency, or convergence, back in 1975. They didn't call it that, but the "Consistency" section sounds a lot like it. According the Lamport, this is the paper that inspired him to write "Time, Clocks, and the Ordering of Events in a Distributed System" because he found flaws in the RFC, which may be my favorite example of https://xkcd.com/386/.
Thanks for the great write-up, Pat! https://jepsen.io/consistency (inspired from Peter Bailis's HAT paper) offers at least one answer to your question on when causal consistency can be an amazing gift, and it's when the system designer is building a "Sticky Available" (available under partition if clients only talk to the same server instead of failing over to another server) system.