README: Don't recommend redis distlock about dsync HOT 4 CLOSED

While Antirez has shown himself to be an excellent C programmer, he's a poor distributed system engineer. Martin Kleppmann is a distributed system researcher at Cambridge. I'm likely to trust his analysis more.

from dsync.

http://antirez.com/news/101 - looks like the jury is out there?

from dsync.

Another common way to implement distributed lock is to simply use Lease.
Xtreemfs use the FLease algorithm that is a Lease built on top of Paxos register: http://www.xtreemfs.org/publications/flease_paper_ipdps.pdf

The main difference I see is that stale lock are automatically released if the server can't renew the lease in time. Also the state change is linearizable and crash resistant.

from dsync.

We are not "recommending" redislock, it is just another approach that may work for some in some cases and not in other cases.

It is not meant to be a conclusive list, for sure there will be other approaches as well.

from dsync.