Re: PROPOSAL: init file actions (draft 2)

["H. Peter Anvin" <hpa@transmeta.com>]

Daniel Quinlan wrote:
> 
> Erik Troan <ewt@redhat.com> writes:
> 
> > I think it would be better for init scripts to call functions which
> > describe what's happening, rather then forcing other bits of init to
> > parse (somewhat) arbitrary text output. This would have the side
> > affect of making init scripts i18n w/o forcing the writer to do any
> > extra work.
> 
> That might work well, except this conflicts with the removal of the
> requirement that init files be scripts.  If init files aren't
> guaranteed to be scripts, then we need separate executable programs
> instead of shell functions.
> 
> (I would still like to specify that init files be /bin/sh shell
> scripts.  I can't recall seeing an init file that wasn't a script, and
> I like giving sysadmins that a little extra bit of flexibility.)

One could argue that you can always wrap whatever you want to use in a
shell script, so as far as I'm concerned, it's probably a noop either
way.  Vice versa, you can always move a binary somewhere else and wrap
it in a script.

One could argue, though, that calling a separate *program* (which may or
may not be a script) rather than a shell function would be cleaner/more
flexible.

	-hpa