I’m still trying to wrap my mind around this algorithm, but it seems like there’s a great deal more messages being passed than there would be in any paxos variants; and the speed will definitely be limited by the slowest participant, where in many paxos variants the speed is limited by the slowest of the fastest N/2+1 participants. So it has some objective drawbacks for some uses, like cross-datacenter or cross-continental quorums where something like epaxos (https://www.cs.cmu.edu/~dga/papers/epaxos-sosp2013.pdf) might be better. Still, very interesting, thanks for the post @vyodaiken.
maybe its obscurity has something to do with it being patented. generally a blocker for algorithms you want to build systems on top of. I stopped reading the post right there.
I would definitely have mentioned that at the top of the article. Because of the broken US software patent system, engineers can’t safely read any information about a patented technology. :(
If you won’t read about patented technologies you commit to not understanding everything from the AC motor to RSA and certainly a great deal of the distributed consensus literature. Good luck with it.
You need to be careful when reading patented stuff (if it hasn’t expired!!!) since infringements have higher penalties if you actually knew the thing was patented. If you read about any patented things, it’s pretty easy to make a jury believe that you infringed willingly, because you actually might have read about the patent in question, etc.
As a result, many people, especially those that work for large corporations, do not read the details about patented technologies.
I’m still trying to wrap my mind around this algorithm, but it seems like there’s a great deal more messages being passed than there would be in any paxos variants; and the speed will definitely be limited by the slowest participant, where in many paxos variants the speed is limited by the slowest of the fastest N/2+1 participants. So it has some objective drawbacks for some uses, like cross-datacenter or cross-continental quorums where something like epaxos (https://www.cs.cmu.edu/~dga/papers/epaxos-sosp2013.pdf) might be better. Still, very interesting, thanks for the post @vyodaiken.
maybe its obscurity has something to do with it being patented. generally a blocker for algorithms you want to build systems on top of. I stopped reading the post right there.
[Comment removed by author]
I would definitely have mentioned that at the top of the article. Because of the broken US software patent system, engineers can’t safely read any information about a patented technology. :(
If you won’t read about patented technologies you commit to not understanding everything from the AC motor to RSA and certainly a great deal of the distributed consensus literature. Good luck with it.
You need to be careful when reading patented stuff (if it hasn’t expired!!!) since infringements have higher penalties if you actually knew the thing was patented. If you read about any patented things, it’s pretty easy to make a jury believe that you infringed willingly, because you actually might have read about the patent in question, etc.
As a result, many people, especially those that work for large corporations, do not read the details about patented technologies.
I am no lawyer but that seems like a hazardous defense. In any case, both this patent and Lamport’s original Paxos patent have expired.
I know that, but it’s not always obvious.