On Mon, 20 Nov 2000, Greg A. Woods wrote:

# [ On Sunday, November 19, 2000 at 18:26:53 (+0700), Robert Elz wrote: ]
# > Subject: Re: Addition to force open to open only regular files 
# >
# > Now I know you have absolutely no idea what you're talking about ...
# > 
# > brandenburg$ ls -lut /etc/passwd; TZ=/etc/passwd date; ls -ltu /etc/passwd
# > -rw-r--r--  1 root  wheel  762 Nov 19 03:26 /etc/passwd
# > Sun Nov 19 10:40:25 GMT 2000
# > -rw-r--r--  1 root  wheel  762 Nov 19 17:40 /etc/passwd
# > 
# > I'm currently at UTC+0700 (ie: 10:40 GMT == 17:40 local).
# > 
# > Now, which file was that that TZ doesn't refer to?
# 
# You didn't show me anything about *ACCESS*.  Just attributes, and one's
# that don't really matter in context since in any context where they
# might matter they are *NEVER* modified by TZ (because if they are then
# the program is buggy by design).

Excuse me.  He *did* show you ACCESS.  Perhaps not modification, but
last I saw, 'ls -lu' shows the time of the last ACCESS (read access)
of the file, directory or recursive directory as listed.

# Sure if some idiot writes a shell script that trys to interpret the
# timestamps as modified by TZ then they'll get in trouble.  But that
# would be a really idiotic thing to do now, wouldn't it.

If they had something that could read a protected file via TZ,
that might not be so idiotic.

# 
# (Which reminds me -- CVS itself has always had this potential problem
# because it does not maniplate timestamps in their basic "time_t offset
# from UTC" format.)
# 
# > For "ping: to not find the same host as "telnet" finds, no matter what
# > the name I use to access it, would be a serious bug IMO.
# 
# Exactly.  That's why leaving name-to-location mapping in the hands of
# the user is a serious security hole and thus why both $HOSTALIASES *and*
# $LOCALDOMAIN must die quickly.

Wow!  Hey, look!  Pinball Tech Support!

				--*greywolf;
--
*BSD, stupid.