In article <12*************@corp.supernews.com>,
go***********@burditt.org (Gordon Burditt) wrote:
I made a test page that just outputs about 100 date('Y-m-d H:i:s');
Occationally (say, 1 out of 30 tries) the time goes backwards by one or
two seconds. Like:
2006-11-14 07:29:16
2006-11-14 07:29:16
2006-11-14 07:29:16
2006-11-14 07:29:14
2006-11-14 07:29:14
2006-11-14 07:29:14
Or even reloading the page, the 2nd page will show an earlier time than
the first page by one or two seconds.
It's no big deal. It's just interesting. I wonder, how can this happen?
Is the server time synchronized to anything? NTP tries to avoid large
jumps like that, but other methods (like maybe "net time" or "ntpdate")
might not.
I have asked my host. It's the same for time() and microtime(), and it
jumps forwards and back, like this:
This is with date() (time()) (microtime()):
2006-11-14 14:30:56 (1163543456) (1163543456.4591)
2006-11-14 14:30:56 (1163543456) (1163543456.4592)
2006-11-14 14:30:56 (1163543456) (1163543456.4593)
2006-11-14 14:30:56 (1163543458) (1163543458.0658)
2006-11-14 14:30:58 (1163543458) (1163543458.0659)
2006-11-14 14:30:58 (1163543458) (1163543458.066)
2006-11-14 14:30:58 (1163543458) (1163543458.0661)
2006-11-14 14:30:58 (1163543458) (1163543458.0662)
2006-11-14 14:30:58 (1163543458) (1163543458.0663)
2006-11-14 14:30:58 (1163543458) (1163543458.0664)
2006-11-14 14:30:58 (1163543458) (1163543458.0665)
2006-11-14 14:30:58 (1163543458) (1163543458.0666)
2006-11-14 14:30:58 (1163543456) (1163543456.4649)
2006-11-14 14:30:56 (1163543456) (1163543456.465)
2006-11-14 14:30:56 (1163543456) (1163543456.4652)
2006-11-14 14:30:56 (1163543456) (1163543456.4653)
2006-11-14 14:30:56 (1163543456) (1163543456.4654)
Those are 1.6 second jumps. It's like there are two clocks.