Re: RFC: New required package: libblkid1
On Thu, Mar 20, 2003 at 12:12:44AM -0600, Steve Langasek wrote:
> The problem of being able to set up machines with read-only root
> filesystems without undue pain, and without causing corresponding pain
> for users of diskless systems. The problem of implementing a filesystem
> layout suitable for the Universal Operating System.
The big question is whether or not read-only root is sufficiently
common case that it's worth torquing things just to support it.
> No, the files it's been suggested that /run be used for do not require
> that it be on the root partition; it just needs to be writable, and
> mounted early. It also does not need to be persistent. If this new
> file is expected to be a persistent cache, it doesn't sound like it fits
> within the parameters of the hypothetical /run.
Yes, the whole point of blkid.tab is that it be persistent, so we
don't have to scan potentially huge numbers of disks to find UUID's or
If the rule is that /run gets nuked at reboot, so it's not persistent,
there will be very few files that that are eligible, and given that
there is also a need for writable, persistent files on the root
filesystem --- again, I ask the question --- what's the point?
> Then perhaps the FHS itself should be thrown out as an idiotic idea,
> given that it codifies a system layout that differs in many aspects from
> traditional Unices.
Actually, I was one of the original people who worked on the FHS, and
I can say that there was a very strong balance between adding features
and not being gratuitously different. Keeping things as much as
possible to other Unix system was a key consideration. Yes, moving
some files from /etc to /sbin was a change, but it other Unix systems
were making the same transition.
So it's not that we can't be different, but there needs to be strong
and convincing advantages towards being different. Given that I don't
believe /run will be able to eliminate the need for persistent,
writable files on the root filesystem, I don't believe /run has
sufficient value to be worth being gratuitously different from other