Yes, because software is famous for being bug free after running for 3 days and never requires a staging environment to match the full production loads…
In any case, the repeated notion that the FAA “just flipped it on” is false. And given the ages and complexity of FAA tech, I’m laughing thinking there’s an available staging environment. :)
Should there be? Of course! But the thought of a decades old federal system\hardware using modern software deployment techniques is laughable. To get to that stage (heh), they’d have to rip everything out, root and stem, start over. And we can’t do that to such a massive, safe and needed system.
Three days is as close to “just flipped it on” as makes no difference with software of any complexity. The conditions that cause the system to malfunction aren’t going to be happening all the time, so you’re going to want multiple cycles of whatever the natural cadance of the system is before you’d feel even moderately reassured that it was going to be stable.
Even just with something with much lower risk, like an e-commerce system, we’d be running enhanced support for at least a week, and through the first few peaks in usage to be able to respond to any hint of an issue like this, and I’ve certainly seen systems overload connectivity that was adaquate for the previous version. It doesn’t happen straight away, but later when a few too many messages need to pass through a link, and start to backlog.
Yes, because software is famous for being bug free after running for 3 days and never requires a staging environment to match the full production loads…
In any case, the repeated notion that the FAA “just flipped it on” is false. And given the ages and complexity of FAA tech, I’m laughing thinking there’s an available staging environment. :)
Should there be? Of course! But the thought of a decades old federal system\hardware using modern software deployment techniques is laughable. To get to that stage (heh), they’d have to rip everything out, root and stem, start over. And we can’t do that to such a massive, safe and needed system.
Three days is as close to “just flipped it on” as makes no difference with software of any complexity. The conditions that cause the system to malfunction aren’t going to be happening all the time, so you’re going to want multiple cycles of whatever the natural cadance of the system is before you’d feel even moderately reassured that it was going to be stable.
Even just with something with much lower risk, like an e-commerce system, we’d be running enhanced support for at least a week, and through the first few peaks in usage to be able to respond to any hint of an issue like this, and I’ve certainly seen systems overload connectivity that was adaquate for the previous version. It doesn’t happen straight away, but later when a few too many messages need to pass through a link, and start to backlog.