Subject: Re: Cloning bdev/cdev devices, step one
To: Eduardo Horvath <eeh@turbolinux.com>
From: Robert Elz <kre@munnari.OZ.AU>
List: tech-kern
Date: 07/13/2000 21:27:13
    Date:        Mon, 10 Jul 2000 10:40:40 -0700 (PDT)
    From:        Eduardo Horvath <eeh@turbolinux.com>
    Message-ID:  <Pine.NEB.4.21.0007101023360.2216-100000@ehorvath>

  | For instance, if you had a machine on a SAN with 1000s of disk and tape
  | devices, you wouldn't want to have a static set of device nodes to
  | identify all of them.

I would.

  | That's 1000 disks * 8 partitions * 2 (raw, block),
  | or 16000 device nodes and dev_t's.

If you have thousands of drives then 16000 inodes & dir entries for the
devices is chicken feed, not even a blip on the radar.

For sure, a flat /dev directory would not be nice to deal with, but that's
all just user space conventions.   The kernel dev_t will continue to work
just fine as a concept (perhaps some of its internal field sizes need
adjustments).

I completely support rationalising the user space code, so that everything
stops assuming /dev/[r]?d0x type names.  Not to prohibit those names, just
to generalise the uses of them so other names can be used, on systems where
that makes more sense.

kre