Alexander E. Patrakov wrote: > Nathan Coulson wrote: >> It sounds like you'd want to have a bootable system even if /usr is >> temporairly not network mountable, so we would need to make the >> console script >> work even without /usr. (This would probably also cover the case where >> / is on a >> separate partition from /usr, but /usr is too corrupted to be mounted).
This is a very important goal. > I failed to express two separate points as two. > > Point 1: moving /usr/share/kbd to /lib/kbd is definitely doable, and it > would allow earlier start of translated messages. This mainly amounts to > the --datadir=/lib/kbd switch and moving binaries from /usr/bin to /bin. > Okay. Very good. Anybody have objections? Can we get this change tested and put into the book? > Point 2: I dislike /usr on network for completely unrelated reason. > Suppose that you have several machines that share one /usr from network. <Snip good advice> I agree with the above completely, however, you forgot about SANs where /usr isn't shared. Granted, I have not put LFS on an enterprise server (Yet!), not to mention that they are usually accessed through dedicated interfaces, provided by the vendor (at least in my limited experience). I believe that we should continue to support remote /usr as long as the changes necessary are manageable. -- DJ Lucas -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
