On Oct 29, 2006, at 11:15 AM, Ingo Schwarze wrote:
Leonardo Rodrigues wrote on Sun, Oct 29, 2006 at 01:45:15PM -0300:
Though, it seems a bit strange that OpenBSD lacks something like
that.
Look at it from a different perspective:
There are other operating systems out there featuring thousands of
lines of complicated scripts just to ensure that users never need
to do simple tasks themselves.
I agree that is overkill for more esoteric tasks but for something as
common as user management the basics should be built into the OS. Why
require users to go writing their own scripts for common tasks? To me
the whole point of using software is to remove the tedious parts and
let the humans concentrate on the parts that the computer cannot do.
In addition to the usual "simplicity improves maintainability and
usability" argument, my impression is that OpenBSD actively encourages
users to understand how the system works - and to understand which
tasks are simple and which ones aren't.
This may seem like simplicity but in reality this forces multiple
implementations for the deluser command as everyone writes their own
scripts. This decreases security since each custom script is unlikely
to be audited for correctness and completeness by as many users as
part of the main source tree would. Additionally, this wastes time by
forcing users to "reinvent the wheel" of user management rather than
adding new features to the OS.
On first sight, an additional option "remove from group" to usermod(8)
might not hurt much. As a second thought, how would you call it, -g
and -G are already occupied; yet it is important for learners to
have option names as few and as mnemonic as possible, and please lets
not get into --remove-from-group. As a third thought, what might be
the next special case that somebody could come up with for plausible
reasons? And finally, once you add an option, you have to live with
it for good, as somebody will certainly rely on it.
The idea that a feature should not be added because we do not have a
handy and simple menmonic for it is absurd. We should consider the
inclusion of a feature on the basis of its usefulness and the number
of users who would benefit. Also, making a feature part of the main
source tree allows the usage and behavior to be standardized. Even if
the perfect mnemonic is already taken as long as the interface is
standardized and sane it will still be usable.
Elio Grieco
