Paxos is so beautiful. It’s a piece of art. It’s way more elegant and beautiful than Raft (as it solves the smallest consensus problem instead of a log). I will die on this hill.
Paxos is so beautiful. It’s a piece of art. It’s way more elegant and beautiful than Raft (as it solves the smallest consensus problem instead of a log). I will die on this hill.
https://www.youtube.com/watch?v=p54W-XOIEF8
[suddenly in a clown costume] "What kind of clown am I, claiming I can make you think better? Why should you pay attention to me?"
[in a suit] "This is not the time to be modest. I have done seminal research in the theory of concurrent and distributed systems for which I won the Turing award. You can stop the video now and look me up on the web." [long pause]
Raft is more understandable, because they wrote a paper that was easy to understand. You are much better off reading Heidi Howard to understand Paxos, than Lamport.
In 2016 some researchers at Stony Brook published the full (Multi-) Paxos algorithm using TLA+, including a couple of nice extras:
https://arxiv.org/abs/1606.01387
The spec is succinct enough to commit to memory and way easier to comprehend than Lamport's prose descriptions (Lamport has never specified the Paxos algorithm in TLA+, although you can find his TLA+ spec for Paxos consensus). The paper also includes a mechanically checked proof and an interesting overview of other related work.