Subject: Re: bin/4167: WIBNI sh supported file/command/etc completion?
To: Jim Wise <>
From: Dustin Sallings <>
List: current-users
Date: 09/28/1997 17:26:46
> On Sat, 27 Sep 1997, Dustin Sallings wrote:
> >      I agree here, /bin/sh is used much more commonly for writing shell
> > scripts than playing the role of an interactive shell in my experience. 
> > Most of the time, it's tcsh or bash for people who want to hang out in a
> > shell. 
> So you're suggesting that our standard user shell should be software
> that is not only not included with the system, but is not compatible
> with our licensing?  The fact that you prefer these shells doesn't mean
> that there aren't many of us who can't stand csh(1)/tcsh(1L) and are
> simply not willing to put up with the bloat and inconsistencies of
> bash(1L)...  Has anyone elso noticed that most of the people who are
> arguing to take important features out of sh(1) aren't _using_ it
> anyways?

	You are very incorrect here.  I user /bin/sh a *lot*.  I use it to
write shell scripts.  I'm trying to avoid the exact bloat you dislike in
bash in the shell that interprets my shell scripts.  *LOTS* of things
commonly start up /bin/sh to do quick interpretations of commands, or scripts,
and none of those things need command history and completion, etc...

> Posix 1003.2 does, in fact, require command line history and editing.
> The suggestion of having fat and thin versions of the shell is valid, I
> suppose (cf Solaris with /sbin/sh (static) and /usr/bin/sh (dynamic)).

	I don't doubt that it's required, but is it required as /bin/sh (as
most shell scripts start with #!/bin/sh).  I would have to believe that a
minimal shell could get the job done faster, and login shells could be in some
other location, that was my point.