Subject: Re: CVS commit: basesrc/usr.sbin/user
To: matthew green <mrg@eterna.com.au>
From: Alistair Crooks <agc@wasabisystems.com>
List: source-changes
Date: 08/27/2002 16:38:17
On Tue, Aug 27, 2002 at 11:49:19PM +1000, matthew green wrote:
> user(8) is a random user process?  LOG_USER is basically for programs
> that non-root people run.  logging user(8) info via it seems *exactly*
> the wrong thing (irrespective of the name being the same) - user(8)
> logs are system logs that should go along side other such information.
> LOG_AUTH seems much more like the right place to me...
> 
> 
> from my quick survey of the source tree, the only programs
> that user LOG_USER are:
> 
> 	vacation(1) - exactly as expected
> 
> 	sesd(8) - without looking, i'd say this should use LOG_DAEMON
> 
> 	libc's _DIAGASSERT() - hi luke!
> 
> 
> LOG_USER is for when some person on my system is writting a program
> that wants to use syslog.  that's a "random user process".  root
> running a program to change user information is decidedly not a
> "random user process" to me.

Just because a process runs with an effective uid of 0 does not mean
that it has to log informational messages in a secure place.  Just
because it has a section 8 man page does not automatically force its
informational messages into a separate file by default.

I prefer to look at the intent behind all this, rather than add
semantics to the openlog(8) man page which aren't already there.

If a person is writing a program which wants to use syslog, they
should be using LOG_LOCAL[0-7], not LOG_USER.

> it occurs to me that chpass(1) & friends should probably do the
> same thing as user(8)...

Good idea.

Regards,
Alistair