On Oct 11, Sven Joachim <svenjoac@gmx.de> wrote: > Rather complex, I'm afraid. Especially as not all architectures even > support an initramfs, AFAIK. I doubt this, since the initramfs can be embedded in the kernel image itself (and indeed it always contains one, it just is empty). But still, then these architectures would not support keeping /usr on a standalone file system, which may be an acceptable compromise. > > Would "mv /bin/* /usr/bin/" and making it a symlink just work, without > > the need to create temporary symlinks in every package as red hat plans > > to do? > Not at all, for there are some packages which ship both /bin/foo and a > symlink /usr/bin/foo to it; those would be broken. Yes, let's assume a slightly smarter algorithm to merge the directories and keep the real binary. What else could go wrong? cd /bin/ for x in *; do [ -e /usr/bin/$x ] && ls -l $x /usr/bin/$x; done cd /sbin/ for x in *; do [ -e /usr/sbin/$x ] && ls -l $x /usr/sbin/$x; done cd /lib/ for x in *; do [ -e /usr/lib/$x ] && ls -ld $x /usr/lib/$x; done -- ciao, Marco
Attachment:
signature.asc
Description: Digital signature