2 ms·
I wonder if package maintainers will be able to use this as a way to test their packages on multiple distributions from a single test system. That would be pret
by ChuckMcM 10y ago
I wonder if package maintainers will be able to use this as a way to test their packages on multiple distributions from a single test system. That would be pretty cool.
I continue to be impressed with the progress the WSL team has made. The usable open source development environment on top of a kernel that can correctly drive all the various peripheral in the box is so very useful to me.
- dredmorbius 10y agoThat's nothing you can't already do with VMs, and avoiding vendor lock-in.
- ChuckMcM 10y agoActually it is different. The difference might not matter in some cases, and in other cases it will. Running on a VM means running on emulated devices, running in a Userland host means running on native devices. WSL is to windows as Wine is to Linux. A really trivial example is the buffer cache. On a Linux kernel it will take big chunks of memory to buffer disk blocks that it has read and dirtied. When run in a VM it continues to have that behavior so for a given disk you have two sets of disk caches, one in the VM and one in the host kernel. Twice the cache with half the management smarts.
- dredmorbius 10y agoI'm not arguing that VMs vs API compatibility layers aren't different, but that the specific testbed case you'd mentioned of emulations on a single host is nothing novel. A compatabilty layer generally takes away the kernel's role. Is that the case here?
- ChuckMcM 10y agoGenerally packages are concerned about interactions in 'user land' and the package manager of the distro. By not carrying around multiple copies of the kernel and its internal data structures this compatibility layer provides a better solution (in terms of resource utilization) for working with several different distributions on a single system than the VM approach does.
- dredmorbius 10y agoSure, but that is something you can already do in a chroot jail.