4 ms·

When I first worked at Apple 30 years ago it was widely acknowledged that R&D had a two-tier system for software development. The rockstar engineers got to work

When I first worked at Apple 30 years ago it was widely acknowledged that R&D had a two-tier system for software development. The rockstar engineers got to work on new stuff and the less productive/creative engineers cleaned up their mess. Bringing new features to maturity was a low-prestige activity and was not rewarded as well. The top engineers joined the hot new projects, shipped the first version, and then moved on to the next big thing. That pattern changed somewhat after Steve Jobs returned, but it still seems to be the case that testing and maintenance are not a very high priority.

draw_down 7y ago
I believe it, because pretty much everywhere is like that in my experience
sixstringtheory 7y ago
Sounds like just about every software job I’ve had over the last 7 years. It sounds like a mgmt problem to me. Either they can’t tamp down the devs who move too fast and break things, or they are driving the devs to do so.
- dhagz 7y ago
  It's a lot harder to demonstrate how the value added by improving stability. Like, what's the difference between 99.93% availability and 99.95% availability? How do you show that increasing test coverage by 5% helps users and brings value to the company? "Hey, so you know how our application is pretty much always working, but will error out once every other blue moon? It does that less now," doesn't sound as flashy as "Look at this thing our app does that it didn't do before!"
  - dmitrygr 7y ago
    > what's the difference between 99.93% availability and 99.95% availability? Quite literally a 40% reduction in crashes
    - saimiam 7y ago
      > 40% A little over 28% since it is a 0’02 reduction over 0’07 non-availability rate.
    - CamperBob2 7y ago
      As so many people have pointed out in this thread, though, crashes are far from the only thing that can keep a user from being able to take advantage of a given feature.
  - pfranz 7y ago
    A few years ago I heard people talking about Apple over-metric-ing. I think even Tim Cook was touting how few kernel panics and crashes there were in modern macOS. They had been tracking crashes and putting effort to reduce them. However, the the new tallest poppy were bugs without metrics, like unexpected behaviors, performance regressions, etc.

draw_down 7y ago

I believe it, because pretty much everywhere is like that in my experience

sixstringtheory 7y ago

Sounds like just about every software job I’ve had over the last 7 years. It sounds like a mgmt problem to me. Either they can’t tamp down the devs who move too fast and break things, or they are driving the devs to do so.

dhagz 7y ago
It's a lot harder to demonstrate how the value added by improving stability. Like, what's the difference between 99.93% availability and 99.95% availability? How do you show that increasing test coverage by 5% helps users and brings value to the company? "Hey, so you know how our application is pretty much always working, but will error out once every other blue moon? It does that less now," doesn't sound as flashy as "Look at this thing our app does that it didn't do before!"
- dmitrygr 7y ago
  > what's the difference between 99.93% availability and 99.95% availability? Quite literally a 40% reduction in crashes
  - saimiam 7y ago
    > 40% A little over 28% since it is a 0’02 reduction over 0’07 non-availability rate.
  - CamperBob2 7y ago
    As so many people have pointed out in this thread, though, crashes are far from the only thing that can keep a user from being able to take advantage of a given feature.
- pfranz 7y ago
  A few years ago I heard people talking about Apple over-metric-ing. I think even Tim Cook was touting how few kernel panics and crashes there were in modern macOS. They had been tracking crashes and putting effort to reduce them. However, the the new tallest poppy were bugs without metrics, like unexpected behaviors, performance regressions, etc.