Subject: Re: standardizing pkg-specific system user/group creation?
To: Tom Spindler <>
From: Alistair Crooks <>
List: tech-pkg
Date: 02/06/2001 10:41:58
On Tue, Feb 06, 2001 at 02:31:54AM -0800, Tom Spindler wrote:
> It appears that there are a couple of different ways that various
> packages create users - postgresql futzes with useradd (which fails
> a good portion of the time on reinstalls); popa3d (which seems to
> work pretty reliably) uses sysutils/user (and seems to have worked
> pretty well for me.)
> Should this sort of thing be standardized? It looks like it was
> discussed on 7 Dec 2000 (under the subject 'useradd is useless
> for creating system users'), but most of the thread was devoted
> to fixing useradd. (which, AFAICT, didn't seem to go much of 
> anywhere.)

There are other packages which do various things to add users, amanda
springing to mind instantly.  That was why addnerd was written in the
first place, from which the useradd(8) grew. Never being able to resist
just one more feature, I added the user(8) frontend. Unfortunately,
user(8) isn't available on Linux or Solaris.

There are other considerations:

1. what happens on systems where users should be added via NIS or NIS+?
2. what happens on systems where users should not be added locally,
but via some Moira-like mechanism?
3. what happens for binary packages?
4. other more mundane stuff which relates to site and company policy

In effect, there isn't really any standard way to do this which will
work across the board.  And, because sysutils/user and the standard
useradd(8) are really the same beast, I'm not sure what's being
suggested here.

But I'm open to suggestions about what should be done.