On Tue, Apr 6, 2010 at 2:27 PM, Paul Prescod <pres...@gmail.com> wrote: > I *believe* that the key messages of those blog posts was: > 1. Using distributed vector clocks are easy once they are implemented. > 2. Implementing distributed vector clocks is hard on the datastore vendor. > 3. If you have long-term network partitions you're kind of screwed (which > is probably true in any case)
I think this is not a problem in practice, where "long term" means hours not weeks. > And if so, is it your plan that "eventually consistent counters" be possible > in a future version of Cassandra? Yes, that is the main reason Digg is working on 580. -Jonathan
