Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!mit-eddie!genrad!decvax!decwrl!pyramid!oliveb!sun!gorodish!guy
From: guy%gorodish@Sun.COM (Guy Harris)
Newsgroups: comp.unix.wizards
Subject: Re: symbolic links  and csh execution
Message-ID: <21681@sun.uucp>
Date: Sun, 21-Jun-87 17:33:47 EDT
Article-I.D.: sun.21681
Posted: Sun Jun 21 17:33:47 1987
Date-Received: Tue, 23-Jun-87 00:49:51 EDT
References: a7879@brl-adm.ARPA> <2211@bunker.UUCP> <1097@mtune.ATT.COM>
Sender: news@sun.uucp
Lines: 59

> Listen guys, if I say 'cd a/b/c/d/e;cd ..' then I am now in a/b/c/d,
> regardless of how many symlinks or networked file systems I had to go
> through to get there. Anything else is brain-damaged.

Which can, in most cases, be done by the shell interpreting the "cd"
command; the question is whether one should add extra stuff to the
kernel to make this work invisibly?  I tend to agree with what was
given as Dennis Ritchie's position, which was "no".

> On a similar note, for the people who are expounding starting one's
> shell scripts with "#! /bin/csh", the logical continuation of this is
> to start one's C programs with "#! /bin/cc", and so on.

No.  Executable files begin with a magic number, which indicates the
interpreter to be used to execute that file.  The "conventional"
magic numbers indicate that the interpreter in question is the
machine's hardware (including microcode if you have it); the "#!"
magic number selects a program to be used as the interpreter.  One
would start C programs with "#! /usr/bin/c_interpreter" or something
like that; unless one has a C interpreter, one does not normally
execute C code.

The intent behind "#!"  was to permit *arbitrary* interpreters to run
programs, not just shells.  (It also permits interpreted programs to
be set-UID, assuming you trust the interpreter's mechanisms to the
degree necessary to make set-UID interpeted programs safe.  Prior to
4.3BSD, there were two very large security holes that rendered
set-UID shell scripts, at least, unsafe; those two holes were closed
in 4.3BSD, but there could well be others.)

> Writing a language processor which is a partial superset of an existing
> standard is one thing, and quite OK in my book; hacking the kernel to
> support this automagically is quite another.

Huh?  How does "#!" "automagically" support "...a language processor
which is a partial superset of an existing standard"?  What it
supports is executing shell scripts as ordinary programs *without*
having to ensure that all programs that might try to execute that
script use "exec[lv]p" instead of "exec[lv]".  The latter is somewhat
nice, considering that not all programs *do* use "exec[lv]p".

> Perhaps csh should have been made to interpret a strict superset of the
> Bourne shell,

Considering that the C shell, I believe, antedated the Bourne shell
(or, at least, the public availability of the Bourne shell), this
might not have been practical.

> so that normal scripts work under it, and then csh scripts should have
> been interpreted via "csh script"? Then we wouldn't have this problem.

"#!" was not concocted merely to solve the C shell vs. Bourne shell
problem.  I presume DMR came up with it as a nice generalization of
the notion of "executable image" and "magic number"; the cost in
terms of kernel code is small, and the benefits seem to outweigh
them.
	Guy Harris
	{ihnp4, decvax, seismo, decwrl, ...}!sun!guy
	guy@sun.com