"In fact, in my experience the process of finding such bottlenecks on running systems is itself quite time-consuming - time which cannot be spent actually reduc

"In fact, in my experience the process of finding such bottlenecks on running systems is itself quite time-consuming - time which cannot be spent actually reducing bloat." Wow. Talk about a self-refuting argument. If it's hard to figure out where to optimize a RUNNING system, how is it supposed to be easier to do so when the system is being created? Arguing "bloat" is "bad for performance"? No problem - graphs or GTFO. Otherwise all you're left with is the emotional "OMG NUMBERS ARE HIGHER NAOW THAN BEFORE" drivel.