Re: LSB specification of runlevels
Theodore Tso writes:
> What's natural is in the eye of the beholder. Certainly "telinit 5"
> is easier to type. You're used to using /etc/init.d/xdm
> {start,stop}, and so it seems more natural to you. But for a novice
> user, and for books for novice users, it's actually much more
> complicated. For example, you need to know whether or not you're
> using gdm, or xdm, or kdm, or some other Desktop Manager. So you
> need to know (and the book needs to explain to you) how to figure
> this out.
>
> Contrast this with "just execute telinit 5 as root".
But it doesn't work everywhere any more than "/etc/init.d/xdm start"
does; specifically, it won't work on Debian (or, presumably, on
Debian-derived systems). A Linux book that says "just execute telinit
5 as root" is not introducing simplicity, it's just wrong.
I think that, from the POV of the inexperienced user, they'd both be
equally regarded as opaque runes.
What I'd *actually* like to see is for the command to be something
like "start-service x-display-manager", for xdm, gdm and all the rest
to register as providing the service x-display-manager, and letting
init sort out the details of getting it going, including automatically
starting any other services required.
Supporting this (supposing some hypothetical standards organization
were to specify it, and it's something that would benefit tremendously
from cross-distribution uniformity) doesn't imply rewriting init, if
you really wanted you could instead just shove all the knowledge into
a shell script which did the right things with /etc/init.d/ and/or
telinit; though one might reasonably think alternative approaches more
elegant.
ttfn/rjk
Reply to: