Because, up until today, windows time did what I needed it to do. It may
still, if the fault turns out to be network related.

In reality, it's more software to learn to administer, and setup and run
on bunch of PC's. As a time nut, I know exactly how much time I need for
all of my other hobbies, since there's never enough of it...  :)

Dan

On 3/25/2013 12:00 PM, time-nuts-request@febo.com wrote:

Dan (I think) writes:

I've been using the standard NTP client in Windows XP for ages, and it
works just fine. I tried third-party stuff. It was just more work for
no apparent gain. The XP desktop is synchronized with my NTP server
perfectly within the limits of my perception, so there is no reason to
go further. Microsecond accuracy is not necessary because I have no
way of testing accuracy with that resolution, nor do I have any
application that requires it. I'm mainly concerned with long-term
accuracy, not short-term accuracy, so it's more important to be
correct within 1/100 second over a period of years than to be correct
within 1 microsecond over a day.

--
Anthony

On Mon, Mar 25, 2013 at 9:05 PM, Anthony G. Atkielski
anthony@atkielski.com wrote:

For most users I think that is reasonable.  It's just not what one
expects to read on  a "Time Nuts" list.  Here we expect to see posting
from true nut-cases who want microsecond just because they can do it.

Gosh what next?  Someone will admit that a normal TTL $2 can
oscillator is actually "good enough" and he is tossing out the GPSDO.

--

Chris Albertson
Redondo Beach, California

Chris writes:

But how can you verify microsecond accuracy on Windows? Even the OS
only has 10 ms resolution for the system clock.

I think there are different ways to be nuts about time. I like time
measurement that is extremely accurate over long periods. Like my
Waveceptor watch, which is accurate to within 1 second in 30 million
years over the long term (because it syncs with atomic time), although
it's free-running accuracy is only about a few hundred milliseconds
per day, and its probably off by a few tens of milliseconds regularly,
due to propagation delay and such.

For me, it's okay if the clock is 10 microseconds off, but it's not
okay if it's off by 10 microseconds today, 20 usec tomorrow, 40 usec
the day after, and so on.

I guess it's the difference between time-of-day measurement and
interval measurement. I'm particularly fascinated by time-of-day
measurement, but much less so by interval measurement. Perhaps because
TOD is easier to verify directly through human perception, whereas
with interval measurement you quickly become dependent on machines for
verification.

--
Anthony

On Mon, Mar 25, 2013 at 10:56 PM, Anthony G. Atkielski
anthony@atkielski.com wrote:

First off, I was hardly serious.  Making fun of someone for being sane?

But the technical answer to your question about how to measure.
Usually with NTP you'd build a set of at least five computers and each
of the five measures the other four.  You will find a core subset that
track each other very well and then measure the Windows system
relative to the local "consensus time."  Being a true "nut" you'd have
have a few GPS receivers to help establish to local time and also a
core set of NTP servers running BSD or maybe Linux.

The thing to remember about NTP is that it is both a server and a
client. So your Windows system will tell the other NTP system son your
network its idea of the correct time.  Those other systems will report
the jitter and offset of the Windows NTP.

Yes it is a lot of work for nothing.

--

Chris Albertson
Redondo Beach, California

On Tue, 26 Mar 2013 05:05:26 +0100, "Anthony G. Atkielski"
anthony@atkielski.com wrote:

I have had trouble with the built in XP NTP client where it fails
silently so I usually install Tardis which keeps an easy to read log
which includes performance data.