It's all about tradeoffs and acceptable risk. A bank, an e-mail provider, and a game analytics tracking system all have different levels of risk tolerance. (Protip: delivery services already lose way more than one in 10 million packages, for any number of reasons.)
I haven't watched the talk yet either, but actually quantifying the risks associated with eventual consistency is a nice idea. It should help system designers make better decisions as to what tools are trustworthy enough for their application.
I haven't watched the talk yet either, but actually quantifying the risks associated with eventual consistency is a nice idea. It should help system designers make better decisions as to what tools are trustworthy enough for their application.