Sysop: | Amessyroom |
---|---|
Location: | Fayetteville, NC |
Users: | 43 |
Nodes: | 6 (0 / 6) |
Uptime: | 107:36:28 |
Calls: | 290 |
Files: | 905 |
Messages: | 76,678 |
1. First, root and ordinary users will not be able to use commands in each other's directories, which will greatly increase their security
1. First, root and ordinary users will not be able to use commands in
each other's directories, which will greatly increase their security
2. If /usr is partitioned separately instead of a unified / partition, ordinary users can also use commands in /usr/bin, which increases convenience
This is not correct. Whether any of /usr/bin,/usr/sbin,/bin or /sbin
share a partition or not has no relationship to whether a user can
invoke a command, or whether that path is searched for unqualified
command names (determined by $PATH).
Hi,
On 12/2/24 18:39, Jonathan Dowland wrote:
This is not correct. Whether any of /usr/bin,/usr/sbin,/bin or /sbin
share a partition or not has no relationship to whether a user can
invoke a command, or whether that path is searched for unqualified
command names (determined by $PATH).
FWIW, I do think that even with user namespaces weakening the distinction between admin and non-admin users, having things you normally need root privileges for not take part in tab completion is still useful to people who use text terminals.
I understand that in the shiny future there are only services that should be started by making an RPC call to a supervisory daemon (so not even root should have them in the path), and tools to make RPC calls (that are potentially useful to any user), but that shiny future has not arrived yet, and in the meantime, all it does is annoy console users.
On 2 Dec 2024, at 09:10, Andrey Rakhmatullin <wrar@debian.org> wrote:
On Mon, Dec 02, 2024 at 09:38:28AM +0800, kindusmith wrote:
1. First, root and ordinary users will not be able to use commands in each >> other's directories, which will greatly increase their security
(typical level of argumentation)
--
WBR, wRAR
1. First, root and ordinary users will not be able to use commands in each >> other's directories, which will greatly increase their security
(typical level of argumentation)
The ability to isolate users from commands that they can’t use anyway is a nice touch, but the kernel and the utilities should handle the first line, IMHO.
Hi,
On 12/2/24 18:39, Jonathan Dowland wrote:
This is not correct. Whether any of /usr/bin,/usr/sbin,/bin or /sbin
share a partition or not has no relationship to whether a user can
invoke a command, or whether that path is searched for unqualified
command names (determined by $PATH).
FWIW, I do think that even with user namespaces weakening the
distinction between admin and non-admin users, having things you
normally need root privileges for not take part in tab completion is
still useful to people who use text terminals.
This is not correct. Whether any of /usr/bin,/usr/sbin,/bin or /sbin share a partition or not has no relationship to whether a user can
invoke a command, or whether that path is searched for unqualified command names (determined by $PATH).
FWIW, I do think that even with user namespaces weakening the
distinction between admin and non-admin users, having things you
normally need root privileges for not take part in tab completion is
still useful to people who use text terminals.
This is not and never was the purpose of /sbin and /usr/sbin. It was a bit
of mistaken confusion when someone at redhat first thought of removing them from users' PATH and represented a kind of lost memory of the history of
this distinction.
/sbin and /usr/sbin were always in everyone's path and should be. There are essential programs that are useful to every user in there. It's incredibly annoying when coming across systems that get this wrong.
The purpose of /sbin and /usr/sbin was to have statically linked essential programs needed to bring up the network, mount filesystems, and do basic admin work in case your large storage device which might require those
tools to mount or repair. This is a necessary feature to achieve a lot of other things like shared /usr volumes and dumb terminals with minimal storage.
But it's absolutely not about user paths or security benefits. (Though
shared mounted /usr actually did have some security benefits). It's about breaking the dependency cycles and being able to start up the system
workout requiring the entire world to be there already.
On Mon, Dec 02, 2024 at 06:46:52AM -0600, rhys wrote:
1. First, root and ordinary users will not be able to use commands in each
other's directories, which will greatly increase their security
(typical level of argumentation)
[...]
It's quite simple, and has nothing to do with security.
You've missed the context, even though you've kept it in your quotes.
Security has nothing to do with it. No user who is a security threat will be thwarted by an inability to specify /sbin manually (or update $PATH).
Read the original email.
1. First, root and ordinary users will not be able to use commands in each
other's directories, which will greatly increase their security
(typical level of argumentation)
It's quite simple, and has nothing to do with security.
Security has nothing to do with it. No user who is a security threat will be thwarted by an inability to specify /sbin manually (or update $PATH).
The ability to isolate users from commands that they can’t use anyway is a nice touch, but the kernel and the utilities should handle the first line, IMHO.1. First, root and ordinary users will not be able to use commands in each >>>> other's directories, which will greatly increase their security
(typical level of argumentation)
My point is that it makes no sense, as unfortunately is typical for these kinds of complaints.
https://css.csail.mit.edu/6.824/2014/papers/plan9.pdf
Of course, these advantages have become impossible now with the
merging of directories
No. Not impossible. One simply has to overcome inertia and the
nostalgia that many leading lights in the Linux community have for 1980s
Unix systems.
As I understand it, the division of root and /usr file systems
originates in the capacity limits of disk packs for the DEC PDP-11 and
VAX-11 series of computers. It's of a piece with the 640kB RAM limit afflicting the IBM PC architecture.
In other words, nonsense that should be discarded with great force at
the first opporunity, which arrived decades ago.