Here's one perspective on naked chroot:
- The only thing virtualized is the filesystem.
- Everything else remains exactly the same. This includes the process tree, memory access, and system calls.
- There are some restrictions on filesystem activity that block the use of features which might allow a surreptitious "breaking out" of the chroot.
Jails add process policy restrictions that either virtualize the chrooted processes, or limit their capabilities. The no-longer-active "sysjail" technique developed for OpenBSD used systrace(4) for this purpose.