alt Hacker NewsThere are very, very strong incentives for performance. Google and other hyperscalers have done studies on their data at scale (and boy do they have a lot of data), and even delays measured in low hundreds of milliseconds harm user retention. On the backend side, 1% improvements in performance can translate to millions of dollars in reduced costs at scale annually. There simply are not enough qualified programmers in the world creating performant software.
With open source it's not even about incentives. I still put effort into the software I make on my own time because I create the kind of software I want to see in the world, ie. software that doesn't feel miserable to use. It's simply about culture. People build up assembly and lower-level abstractions in general to be the scary monster in their closet, and not something they could actually learn if they just tried.
> There are very, very strong incentives for performance. Google and other hyperscalers have done studies on their data at scale (and boy do they have a lot of data), and even delays measured in low hundreds of milliseconds harm user retention. On the backend side, 1% improvements in performance can translate to millions of dollars in reduced costs at scale annually.
for Google and other hyperscalers, not for mom and pop shops and electron apps.
> There simply are not enough qualified programmers in the world creating performant software.
Nonsense. You seriously think there’s some arcane knowledge in optimizing things? Sure, if you’re pushing microseconds and optimizing network stack just to squeeze last drops out of it. But majority of software runs stupid quadratic loops, overuses map/filter/reduce, instantiates too much and is bloated with useless features. It takes one capable programmer to optimize this mess to roughly 90-98% of what’s possible. It takes world class to squeeze last 2%, but majority of software doesn’t need or care about it.