News for hackers. Stuff that matters.
Re:First private Chinese attempt to send rocket into space f... by dhruval on Sunday October 28, 2018 @03:32AM (#18319532)
Rockets are very hard and failure is very costly. Requires a different mindset from software. All the best to the team.
Re:First private Chinese attempt to send rocket into space f... by baybal2 on Sunday October 28, 2018 @03:43AM (#18319565)
>Requires a different mindset from software. All the best to the team.

Definitely - it is all about the software team forgetting the accursed word "iteration"

There, the software must works right from first compile. The way of writing software "in hopes it works," and reworking it by going from one newly discovered roadblock to another doesn't work.

Also, formal verification is genuinely needed here, even more than say in embedded space.

Re:First private Chinese attempt to send rocket into space f... by seanmcdirmid on Sunday October 28, 2018 @10:46AM (#18320805)
You do have simulations in place, so iteration and testing are still very important. Formal verification, when used at all, is still not a substitute for lots of testing.
Re:First private Chinese attempt to send rocket into space f... by nicois on Sunday October 28, 2018 @03:58AM (#18319601)
My naive perspective is not that software engineers should forsake iteration; rather that their tests / simulations have to be "out of this world". After all, the software is still written by taking real world constraints and models and then applying them in software. Anything you are counting on software to do properly should be testable, up to a point, in a virtual environment.
Re:First private Chinese attempt to send rocket into space f... by baybal2 on Sunday October 28, 2018 @04:09AM (#18319631)
>rather that their tests / simulations have to be "out of this world".

Formal verification far exceeds testing as know to mainstream software development.

And in addition to it, some "sanity tests" are preferably to be done just to have an assurance that your are not verifying the wrong thing. I heard of few cases when formally verified software failed despite devs being religiously zealous with it.

One on my mind was of oxygen sensor wired to ADC, and ADC was able to measure negative voltage... Thus it was keeping the fuel injector running for the negative amount of time... You can imagine the following sequence of events.