Chas Emerick countered: “No matter how well you know the landmines, imprecise floating point is a hazard.
You can substitute “manual memory management” for “floating point” in your third paragraph, and it remains a plausible statement — which means to me that floating point error is worth worrying about (along with modeling and approximation error):
‘Manual memory management can be tricky. Things can go spectacularly bad for reasons that catch you off guard if you’re unprepared. But I’ve been doing programming long enough that I believe I know where the landmines are and how to stay away from them. And even if I’m wrong, I have bigger worries.’ “
Interesting given the recent discussions on memory safety.
That doesn’t tell us floating point is worth worrying about, just that that particular argument does not show it’s not worth worrying about. To see this, replace “manual memory management” with “Python default function arguments”.
The argument for why floating point is worth worrying about is its use in safety- and mission-critical systems. People worrying about it leads to research on tools to detect the errors. A percentage of them will be open source. Once they mature, we trade our worries for extra coding guidelines and another analysis pass in the CI system. Those tools are in development right now, too. :)
I think that’s part of the perception mismatch: John Cook does simulations and modelling, not finance or safety. So in his domain, it’s not clear that floating point error is the most dangerous kind of error- certainly not the one statisticians should be most worried about.
Chas Emerick countered: “No matter how well you know the landmines, imprecise floating point is a hazard.
You can substitute “manual memory management” for “floating point” in your third paragraph, and it remains a plausible statement — which means to me that floating point error is worth worrying about (along with modeling and approximation error):
‘Manual memory management can be tricky. Things can go spectacularly bad for reasons that catch you off guard if you’re unprepared. But I’ve been doing programming long enough that I believe I know where the landmines are and how to stay away from them. And even if I’m wrong, I have bigger worries.’ “
Interesting given the recent discussions on memory safety.
That doesn’t tell us floating point is worth worrying about, just that that particular argument does not show it’s not worth worrying about. To see this, replace “manual memory management” with “Python default function arguments”.
Yes, it just calls out the bad argument.
The argument for why floating point is worth worrying about is its use in safety- and mission-critical systems. People worrying about it leads to research on tools to detect the errors. A percentage of them will be open source. Once they mature, we trade our worries for extra coding guidelines and another analysis pass in the CI system. Those tools are in development right now, too. :)
I think that’s part of the perception mismatch: John Cook does simulations and modelling, not finance or safety. So in his domain, it’s not clear that floating point error is the most dangerous kind of error- certainly not the one statisticians should be most worried about.
Makes sense.