On Tuesday 2016-05-24 00:46, Ole Tange wrote: >Should GNU Parallel ignore, kill or wait for background children?
Implementations that are capable of tracking all children ever created (such as cgroups, as for example used by systemd) have no reason to ignore, since they know what processes to wait for/to kill. Whether wait or kill, that's best left as a user option. Even programs not as controversial as init systems show different default behavior when it comes to (all kinds of) child "objects". ssh -L would wait for all tunneled TCP connections to close first, while ending the initial X client that was started from xinitrc will cause progression in the xinitrc script and often tear down the X server, killing everything else connected to it.
