On Thu, May 29, 2014 at 1:53 PM, Gary Thompson <gthompson@xxxxxxxxxxx> wrote:
All of this by my assumption, from reading, that other systems may update
Changed date on access alone.
If you got that from me, then you probably misread the following:
From: MIDRANGE-L [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of John Yeung
There's a system call named "stat" on both Unix and Windows, that
retrieves three times. "Access" time and "content modification"
time have the same meaning on both systems,
The above is saying that
(1) The meaning of "access" time on Unix is the same as the meaning of
"access" time on Windows.
(2) The meaning of "content modification" time on Unix is the same as
the meaning of "content modification" time on Windows.
That was to contrast with the third time, which has a different
meaning on Unix than it does on Windows.
I keep saying "the third time". What I really mean is the "ctime".
The access time and modification time are named "atime" and "mtime",
respectively. I really wanted to avoid bringing up actual names used
in implementation, just as I try to avoid using actual DDS field names
when talking to users about what's in the database. But in this case,
maybe knowing the raw names sheds some light on things. For the
morbidly curious (or bored), here is a nice article on the "stat"
system call (focusing on Unix, which is where it originated):
There is even a section in there specifically about ctime.
But the main takeaway that I was going for was that timestamps are
fundamentally unreliable, on any computer system whatsoever. Your
best bet is, if possible, to use techniques which don't depend on
timestamps at all. Some ideas for doing that were mentioned by others
in this thread.