Comment by giulivo
13 years ago
Given the topic of the article I understand my question could be seen as OT but I'd like to dig more into the following:
> The NT kernel is still much better than Linux in some ways --- you guys be trippin' with your overcommit-by-default MM nonsense
What's wrong with the overcommit-by-default behaviour? Or how/why the NT kernel is supposed to do better?
I don't know what the author was thinking of specifically, but the OOM killer seems like a bit of a hack to me. I can think of situations where I'd rather have an allocation fail upfront predictably than be randomly killed by the system.
A lot of people seem to think you're screwed when you're out of memory and so think killing the process is acceptable, but I've worked on code bases where it's actually handled somewhat gracefully. (Although, on NT, if you run out of nonpaged pool you start to get weird, random I/O failures.)
but again, if it is just about the default behaviour which we debate, one can simply disable the overcommit
vm.overcommit_memory = 2
see https://www.kernel.org/doc/Documentation/vm/overcommit-accou...
And that was exactly the criticism, wasn't it? "Overcommit-by-default nonsense". I didn't write the original, just trying to parse it. (And with that aspect I kind of agree, it's a goofy behavior.)
I guess there's another, somewhat social aspect of making that the default: everyone else coding on the platform assumes the behavior is the other way, so I suspect when you set that flag to off, your other software starts crashing on null pointer dereferences because no one thinks malloc can fail.