On Mon, Mar 15, 1999 at 11:47:24AM -0600, Rob Browning wrote: > The reasons we have the / /usr split are historical and practical. > Back in the early days there was a small amount of fast "disk", and a > larger amount of slow disk. On one of the early systems, I believe > the latter was actually a large rotating drum. You wanted to have the > stuff that needed to be faster in /. Today, practically speaking, you > want / to be small and separate from /usr so you can hopefully limit > the likelihood that a critical failure will hit the tools you need to > deal with a critical failure. AFAIR /usr is separate so you can mount it read only. At least this it what I do. As /usr is the biggest partition on my workstation I would gain a lot of time because it need not be check after a crash. Sadly the machine did not crash since I set it up (last year)... ;) > >From what I've heard this becomes irrelevant in the HURD because it > has a much more sophisticated filesystem. You can "interleave" mount > points, so one partition can provide some of the files in /bin, and > another partition can provide the rest (once it's mounted). Perhaps > something like: > > mount /dev/foo1 / # For normal root files. > mount /dev/foo2 / # add normal /usr files. > > Though I have no idea how this is actually set up, and I don't know > how potential name conflicts are resolved... What about WRITING to this mount point? How should the installation be done? mount /dev/foo1 / [install base system] mount /dev/foo2 / [install the rest] Nice. What about apt-get dist-ugprade? cu Torsten
Attachment:
pgppUZpjMipKB.pgp
Description: PGP signature