out of memory
When a resource is exhausted, like memory, does this need special handling in a programming language? I see a way to avoid it, thus answering no. But I wonder what kind of argument requires yes instead of no. Note I rarely find paper references either useful or interesting, since I like basic ideas summarized in granularity of dozens to hundreds of words.
Resource allocation can look like a system call to a PL (programming language), so running out of memory can be handled by the environment. A runtime hosting a PL can react instead. For example, in a lightweight process system, max memory used by one process might be limited. Hitting a limit occurs in a call, blocking the group/process/thread/fiber. What happens next depends on configuration, process architecture, and registered preference. Options include killing a process group, relaxing limits in triage, and/or blocking until space is freed by another scheduled entity under the same limit-scope mechanism.
An "important" process ought to have an out-of-memory handler, to react more flexibly than letting that process be killed. In a lightweight process system, this can be a signal awakening a fiber handling high priority event notifications like "you're overdrawn". From a PL perspective, the code is just event handling, like any other event. A runtime or operating system sees a resource exhausted, so a scheduler responds and not the PL. Is there a reason why a PL should get any say?
I can imagine a PL resisting management of resources by the OS and/or runtime because there's no concept of running inside bounded space, so limits imposed by the environment fail to be enough room all the time, because a PL can't stay inside and handlers basically fire all the time to no productive result. But that seems like a language arrogating to itself some OS resource handling role without suitable responsibility and competence. Failing to live inside limits seems more flaw than feature.
My point of view isn't interesting to me because I understand it. I'd prefer to hear different views so I learn something. Is there a reason why a PL provides special value by handling resource exhaustion more directly than the host environment?
Active forum topics
New forum topics