It seems to me that if the reordering client cannot contact other clusters then the system is essentially unavailable. On that basis I don’t think this has really solved the problem it thinks it has.
Indeed; this seems like a rehashing of vector clocks, but now as a panacea. I love vector clocks, but they too have their drawbacks and good points. You’ve highlighted an obvious one.
It seems to me that if the reordering client cannot contact other clusters then the system is essentially unavailable. On that basis I don’t think this has really solved the problem it thinks it has.
Indeed; this seems like a rehashing of vector clocks, but now as a panacea. I love vector clocks, but they too have their drawbacks and good points. You’ve highlighted an obvious one.