4 ms·

As the author also found out (but it's mentioned only later in the post), there is already a Linux system call emulator, "linuxemu", written by Russ Cox and Cin

As the author also found out (but it's mentioned only later in the post), there is already a Linux system call emulator, "linuxemu", written by Russ Cox and Cinap Lenrek: http://9p.io/wiki/plan9/Linux_emulation/index.html http://9p.io/wiki/plan9/Linux_emulation/index.html It seems to be rather outdated by now (only supports 32-bit x86 and might be problematic with current binaries), but it's probably a less resource-intensive solution than integrating a VM. Of course, this depends on the kind of Linux applications you intend to run, emulating an X11 or Wayland environment probably won't make too much sense.

lproven 3y ago
[Author/submitter here] Yes, I did find out in the responses to the post -- which was part of the point of the exercise. Yes, there is indeed, but as you say, it's neglected, it's 32-bit only, and updating it would probably be a substantial amount of work. It was for the one in openSolaris, which is one of the jobs that Joyent did in making SmartOS. Then, if it were updated, it would need continuous maintenance to keep parity with the Linux kernel. It's a moving target, and an exceptionally complex and difficult moving target as well. Whereas what I'm suggesting -- and that's all it is, a suggestion, not even a proposal -- needs neither, it provides a much higher level of compatibility for little work, and merely by updating the kernel in the VM you get 100% Linux compatibility essentially for nothing. The FreeBSD bhyve hypervisor does sparse memory allocation for guests: only guest RAM that's used is allocated on the host. So the guest instances could remain pretty small. An unstated assumption is that this is an idea for 64-bit OSes (host and guest both). 32-bit Linux is on the way out now anyway. It might use more computer resources but the idea is that it uses less programmer resources, which is a large part of the point.

lproven 3y ago

[Author/submitter here] Yes, I did find out in the responses to the post -- which was part of the point of the exercise. Yes, there is indeed, but as you say, it's neglected, it's 32-bit only, and updating it would probably be a substantial amount of work. It was for the one in openSolaris, which is one of the jobs that Joyent did in making SmartOS. Then, if it were updated, it would need continuous maintenance to keep parity with the Linux kernel. It's a moving target, and an exceptionally complex and difficult moving target as well. Whereas what I'm suggesting -- and that's all it is, a suggestion, not even a proposal -- needs neither, it provides a much higher level of compatibility for little work, and merely by updating the kernel in the VM you get 100% Linux compatibility essentially for nothing. The FreeBSD bhyve hypervisor does sparse memory allocation for guests: only guest RAM that's used is allocated on the host. So the guest instances could remain pretty small. An unstated assumption is that this is an idea for 64-bit OSes (host and guest both). 32-bit Linux is on the way out now anyway. It might use more computer resources but the idea is that it uses less programmer resources, which is a large part of the point.