bad practice: not reading the manpage

[Matthew Miller]

On Mon, Jun 13, 2005 at 08:32:23PM -0700, Florin Andrei wrote:
> > 'off' - configure the service to not start
> > 'del' - remove all state for the service entirely
> But that's a state in itself, functionally equivalent to "off on all
> runlevels". It's unique only in the way that it is not preserved by "rpm
> -U"

How *could* it be??? That's the _exact_ difference between using "--del" and
"off".

> Making --list easier to read is actually very valuable. If you have a
> lot of things to take care of, you don't want to spend "brain cycles"
> needlessly; plus, which services are on or off is a pretty important
> issue, you don't want to make mistakes there.

This is a different issue than whether state information should be left on
the system by a command named "--del". There would be several ways to make
the output of this command more readable without mucking with functionality.
(Like, dealing with that nasty xinitd kludge.)



> I understand, "--del" takes the service to a state that's not preserved
> by rpm. Fine. But who did --del? Most likely the sysadmin. Why should
> the software disregard a human decision? The system did not end up in
> that state at random, but by human intervention. The software should not
> overrule it.

The sysadmin shouldn't do that, really, without *expecting* it to get reset
to the default at some point -- or, at least, going "oh, duh" when it does.

> Why should --del be different, other than "this is the way our
> forefathers did it"?

Um, because *removing state information is what --del does*!

There are two possible alteratives: 1) make *no* services (including syslog,
crond, keytable, iptables, etc) start by default or 2) have very
inconsistant and basically unpredictable behavior between upgrading and
installing a package. 

Personally, I think it's *much* better to leave things working as they are
-- and to teach people to use the easier and shorter off/on commands.

> This is the core of the problem and I don't think I received a good
> answer yet.

You have now. :)

> Tradition is fine and all that, but it should change when it's hampering
> the usability. Maybe I spent too much time lately with the Gnome HIG and
> stuff like that (not strictly related, I know, but you get the idea),
> but I think it's the computer semantics that should bend over backwards
> to adapt to human semantics, not the other way around.

Are you seriously suggesting that "--del" is more human-friendly than "off"?

-- 
Matthew Miller           mattdm at mattdm.org        <http://www.mattdm.org/>
Boston University Linux      ------>                <http://linux.bu.edu/>
Current office temperature: 81 degrees Fahrenheit.