Excepting kernel scheduling, dealing with the set of problems i,j,k and being interrupted to deal with the set of problems x,y,z means all your short-term memory dealing with i,j,k disappears, and whatever mental work you’ve done to deal with i,j,k needs to be repeated once x,y,z is “problem solved” - but that’s a waste of time, compared to solving i,j,k completely and then moving on to x,y,z, and it’s annoying for the person solving the problems
I agree, it’s not usually a good use of time to reinvent the wheel, but what if there are unacceptable (e.g.) security risks with all the existing solutions? Then “just right” means “adequately addresses realistic security risks, where core fails to do so.” Developers correctly focus on building things, these aren’t often solid from a security perspective, and sometimes that matters. Abstracted, my point is that your point is correct in one domain, and there’s probably always another, peripheral domain where that’s not true: underlying assumptions mean whole ecosystems are flawed when used for a specific purpose: security, high availability, parallelism, etc
- “10x developers” require 10x pay - this is wanky as shit