> Saying that sacrificing uptime is a good thing for achieving a particular goal is a take that's so wildly out of line with what most of the industry believes

> Saying that sacrificing uptime is a good thing for achieving a particular goal is a take that's so wildly out of line with what most of the industry believes is refreshing, because it really makes you think. What would the world look like, if instead of being obsessed with SRE we simply assumed or accepted that there will be downtime, and treated it like something perfectly normal. Maybe a nice message at ingress/load balancer level, telling everyone that the site will be back in a bit. The difference is that here, the downtime saves energy. "Normal" downtime is unintentional. Servers are still running, just in an error state. They may be constantly attempting to restart themselves at some level (looking at you, k8s). Users are still trying to hit them and their requests may be partially handled. Traffic may be pushed over to newly spun up instances. We can accept that things happen and running large internet services is hard, and SREs and developers everywhere will rejoice, but it won't save energy.