Sticky bit?

[Len Reed]

In article <1991Mar24.223426.10401 at cs.umu.se> dvldbg at cs.umu.se (Daniel Brahneborg) writes:

>The sticky bit is used for changing the user-id of the process that
>runs the program, such as 'su', and lots of others.

I'll be charitable and assume that this was a mental slip and not really
and indication that you don't know what you're talking about.  What
you say is certainly wrong.  You've just described the set-uid bit,
though a complete explanation of that bit would require that you say
what the user id is changed to and a brief description of effective
versus real user ids.

I don't understand why persons post erroneous answers to simple questions.
Again, perhaps this wrong answer was a mental slip, but others have
posted wrong answers to this question that show, without a doubt,
they don't know what they're talking about.

The sticky bit goes back to at least edition 6 Unix; we're talking the
1970s here.  In a classic sense it applicable only to executables--not
even shell scripts.  It can be set only by the superuser.  It says
that when the last process executing this inode dies, do *not* discard
the sharable text from the swap area, but keep it in the swap area.  When
someone later exec's the program, the kernel doesn't have to load the text
from the file, only the initialized data.  This allegedly speeds startup
of progams that are used frequently (e.g., vi).  It has no effect on 
programs that the shell that tend to always have one
or more processes running them.

The meaning is expandable to systems with paging virtual memory: retain
the sharable pages.

The sticky bit (and the setuid and setgid bits, for that matter) have
been appropriated for various uses on BSD and perhaps other systems;
others have written about these uses.  My description of the sticky
bit is the only one applicable to "Unix" in general.
					 
-- 
Len Reed
Holos Software, Inc.
Voice: (404) 496-1358
UUCP: ...!gatech!holos0!lbr