Hi Folks!

We are proud to announce that now more than 5000 people downloaded our
Windows Installer for NTP from www.meinberg.de/english/sw/ntp.htm !

Only a few of you reported problems which we were able to resolve by
providing patched versions of NTP (especially the NT users faced
difficulties) mostly. If you experience problems with the installer,
contact us at ntpsupport@meinberg.de, maybe we can help you.

I'm going to prepare a new version of the installer including a recent
NTP version and would like to invite everyone to let me know what should
be changed/added. Please use the ntpsupport@meinberg.de address for your
suggestions, too.

Thanks to all of you who already sent feedback and/or comments.

Kind regards,
Heiko

In article <4u6fq2-0kv.ln1@gateway.py.meinberg.de>,
 Heiko Gerstung <heiko.gerstung@meinberg.de> wrote:

> Thanks to all of you who already sent feedback and/or comments.

I just gave it a shot, but after running it for about half an hour from 
known-good servers it was still 108 seconds off (even though the Win32 
time service kept it synced properly[1]). It was >1min slow compared to 
my WWVB clock on the wall here.

Unfortunately, I don't have that much time (har har) to poke around with 
it right now, so I uninstalled it (but kept the installer handy for 
later experimenting). Does the uninstaller script automatically undo all 
the changes that were made while installing (i.e. creating the "ntp" 
user, re-enabling the Win32 time service, etc.)?

Looks like a good piece of software to have. I have an old Cobalt RaQ3 
that is serving time right now, but I've been having a remarkable amount 
of difficulty upgrading it to a new, updated version of Linux that will 
run a modern version of ntpd. I have a Windows machine that's on all the 
time here, which would make an excellent server in addition to my Mac OS 
X machine which is used as my main workstation (and is thus unsuitable 
for such services).

I'll definitely have to play with this a bit more. Thanks!

[1] Within the limits of the Win32 time service checking every 7 days.

-- 
Pete Stephenson
HeyPete.com

Hi Pete,

thank you for your comments and feedback, I really appreciate that.

Pete Stephenson wrote:
> In article <4u6fq2-0kv.ln1@gateway.py.meinberg.de>,
> [...] I just gave it a shot, but after running it for about half an hour from 
> known-good servers it was still 108 seconds off (even though the Win32 
> time service kept it synced properly[1]). It was >1min slow compared to 
> my WWVB clock on the wall here.

Mmmh, would be interesting to see your stats (e.g. ntpq -p output and
stuff).

> Unfortunately, I don't have that much time (har har) to poke around with 
> it right now, so I uninstalled it (but kept the installer handy for 
> later experimenting). Does the uninstaller script automatically undo all 
> the changes that were made while installing (i.e. creating the "ntp" 
> user, re-enabling the Win32 time service, etc.)?
I decided not to delete the ntp user as I cannot find out if something
else is relating on the existence (you might have created not only a ntp
user but one who is used for several background tasks and the installer
cannot find out). Re-enabling the time service sounds like a good idea
(better than using no synchronization at all), I could add a dialog
showing up at uninstall time: "During installation the W32 Time Service
has been disabled. Do you want to turn it on again? [YES] [NO]" ...

> Looks like a good piece of software to have. I have an old Cobalt RaQ3 
> that is serving time right now, but I've been having a remarkable amount 
> of difficulty upgrading it to a new, updated version of Linux that will 
> run a modern version of ntpd. I have a Windows machine that's on all the 
> time here, which would make an excellent server in addition to my Mac OS 
> X machine which is used as my main workstation (and is thus unsuitable 
> for such services).

As we all know (do we?), Windows is not the best OS to be used for time
keeping, but I guess you do not need sub-second accuracy, so why not?

> I'll definitely have to play with this a bit more. Thanks!

Yes, check it out. I love to receive more feedback from all of you.

Kind regards,
Heiko

In article <42d58332$0$18644$14726298@news.sunsite.dk>,
 Heiko Gerstung <heiko.gerstung@meinberg.de> wrote:

> I decided not to delete the ntp user as I cannot find out if something
> else is relating on the existence (you might have created not only a ntp
> user but one who is used for several background tasks and the installer
> cannot find out). Re-enabling the time service sounds like a good idea
> (better than using no synchronization at all), I could add a dialog
> showing up at uninstall time: "During installation the W32 Time Service
> has been disabled. Do you want to turn it on again? [YES] [NO]" ...

Fair enough. After a reboot, it would seem that the Win32 time service 
has been re-enabled.

> > Looks like a good piece of software to have. I have an old Cobalt RaQ3 
> > that is serving time right now, but I've been having a remarkable amount 
> > of difficulty upgrading it to a new, updated version of Linux that will 
> > run a modern version of ntpd. I have a Windows machine that's on all the 
> > time here, which would make an excellent server in addition to my Mac OS 
> > X machine which is used as my main workstation (and is thus unsuitable 
> > for such services).
> 
> As we all know (do we?), Windows is not the best OS to be used for time
> keeping, but I guess you do not need sub-second accuracy, so why not?

Correct. I would very much like to afford a small, 1U-2U system that I 
could transfer the processes running on the RaQ to. Mostly it does 
webhosting, and has the built-in control panel which makes it very easy 
for the few friends I host for to administer their own accounts. I'm not 
familiar enough (nor have I done sufficient research into) the various 
open-source control panels and whatnot.

A FreeBSD system would be optimal for timekeeping, at least according to 
information I've seen posted here...but again, I lack time and money. 
Recreational and minor competitive shooting has been taking up most of 
my money, while various academic courses have been occupying my time.

> > I'll definitely have to play with this a bit more. Thanks!
> 
> Yes, check it out. I love to receive more feedback from all of you.

My pleasure. I'll be sure to send you any more feedback I have.

Cheers!

-- 
Pete Stephenson
HeyPete.com

Pete Stephenson wrote:
> Fair enough. After a reboot, it would seem that the Win32 time service 
> has been re-enabled.
> 
That's bad. That means that the Win32 time service has not been 
disabled, just stopped. A disabled service won't be started on boot 
while a stopped one on automatic will restart.

Heiko, you will need to fix that.

Danny
_______________________________________________
questions mailing list
questions@lists.ntp.isc.org
https://lists.ntp.isc.org/mailman/listinfo/questions

In article <42D5C900.2010200@gis.net>, mayer@gis.net (Danny Mayer) 
wrote:

> That's bad. That means that the Win32 time service has not been 
> disabled, just stopped. A disabled service won't be started on boot 
> while a stopped one on automatic will restart.
> 
> Heiko, you will need to fix that.

Perhaps not.

Recall that I /uninstalled/ the NTP package and restarted. If the NTP 
uninstaller re-enabled the Win32 time service when the NTP software was 
removed, this is normal.

Unfortunately, I did not attempt restarting after installing the NTP 
software -- I only did so after I removed it to ensure the Win32 time 
service came back up.

-- 
Pete Stephenson
HeyPete.com

Danny,

Danny Mayer schrieb:
>> Fair enough. After a reboot, it would seem that the Win32 time service 
>> has been re-enabled.
>>
> That's bad. That means that the Win32 time service has not been 
> disabled, just stopped. A disabled service won't be started on boot 
> while a stopped one on automatic will restart.
> 
> Heiko, you will need to fix that.
It is of course disabled when NTP is installed. We were talking about 
re-enabling W32Time when someone chooses to uninstall NTP (this is 
against my religious beliefs, but as I said maybe it's better to have a 
simple time sync service than none at all). I will implement a dialogue 
asking the user if she/he wants to re-enable W32time when uninstallation 
of NTP has been completed.

We're safe, I guess ;-)

Kind regards,
Heiko



> 
> Danny
> _______________________________________________
> questions mailing list
> questions@lists.ntp.isc.org
> https://lists.ntp.isc.org/mailman/listinfo/questions
> 


-- 
Meinberg radio clocks: 25 years of accurate time worldwide

MEINBERG Radio Clocks
www.meinberg.de

Stand alone ntp time servers and radio clocks based on GPS, DCF77 and 
IRIG. Rackmount and desktop versions and PCI slot cards.

Heiko Gerstung wrote:
[]
> As we all know (do we?), Windows is not the best OS to be used for
> time keeping, but I guess you do not need sub-second accuracy, so why
> not?

A lightly loaded Windows system can easily achieve sub-100ms accuracy:

Windows 2000 1GHz Pentium III system:
  http://www.david-taylor.myby.co.uk/mrtg/hermes_ntp.html

Somewhat older Windows NT4 AMD 266MHz system:
  http://www.david-taylor.myby.co.uk/mrtg/bacchus_ntp.html

The systems doing heavy CPU, disk etc. work have more glitches (at least 
the Windows 2000 system [Stamsund] does!)
  http://www.david-taylor.myby.co.uk/mrtg/daily_ntp.html

Cheers,
David 

Pete,

Pete Stephenson wrote:

> In article <42d58332$0$18644$14726298@news.sunsite.dk>,
>  Heiko Gerstung <heiko.gerstung@meinberg.de> wrote:
> 
>> I decided not to delete the ntp user as I cannot find out if something
>> else is relating on the existence (you might have created not only a ntp
>> user but one who is used for several background tasks and the installer
>> cannot find out). Re-enabling the time service sounds like a good idea
>> (better than using no synchronization at all), I could add a dialog
>> showing up at uninstall time: "During installation the W32 Time Service
>> has been disabled. Do you want to turn it on again? [YES] [NO]" ...
> 
> Fair enough. After a reboot, it would seem that the Win32 time service
> has been re-enabled.

If you have observed this on your system I think you have changed the NTP
installation options.

I've also run the installer ntp-4.2.0a%401.1354-o-win32-setup.exe with
default settings on a test machine and after the setup has finished I find
that the w32time service is stopped, and its autostart setting has been
changed to "disabled".

That version of the NTP installer does not re-enable w32time if NTP is
uninstalled, so I assume that your w32time service has been running all the
time. That ight also explain why NTP has been unable to discipline your
system time, as you have mentioned in an earlier post.


Martin
-- 
Martin Burnicki

Meinberg Funkuhren
Bad Pyrmont
Germany

In article <b7jhq2-fu7.ln1@gateway.py.meinberg.de>,
 Martin Burnicki <martin.burnicki@meinberg.de> wrote:

> That version of the NTP installer does not re-enable w32time if NTP is
> uninstalled, so I assume that your w32time service has been running all the
> time. That ight also explain why NTP has been unable to discipline your
> system time, as you have mentioned in an earlier post.

Possibly, but it said something about "RPC Service Disabled" or 
something along those lines when I opened the "Internet Time" panel on 
the system clock...it's 1:30am and I saw it on the screen for a few 
moments several hours ago, so I don't remember the exact message.

I'll play around with it this weekend.

-- 
Pete Stephenson
HeyPete.com

In article <igbhq2-67m.ln1@gateway.py.meinberg.de>,
 Heiko Gerstung <heiko.gerstung_removeme_@meinberg.de> wrote:

> Meinberg radio clocks: 25 years of accurate time worldwide

Heiko,

I'm curious how much your LANTIME/GPS product costs. Would you mind 
contacting me by email (my address is valid). Just curious...I don't 
really have any money now, but I'm always looking for new toys. :)

-- 
Pete Stephenson
HeyPete.com

David,

There is no need to add 100 milisecond to the offset value. You may
modify the script to generate positive values for both positive and
negative offsets and plotting them on mrtg using both channels (I and
O). In this way you may see a more precise graph.

Anyway, on large periods the "Average" value reported by mrtg MUST be
near 0 miliseconds (at least 1 Day Average Value).

Eugen COCA wrote:
> David,
>
> There is no need to add 100 milisecond to the offset value. You may
> modify the script to generate positive values for both positive and
> negative offsets and plotting them on mrtg using both channels (I and
> O). In this way you may see a more precise graph.
>
> Anyway, on large periods the "Average" value reported by mrtg MUST be
> near 0 miliseconds (at least 1 Day Average Value).

Thanks for that, Eugen.  Yes, I know that there are different plotting 
routines to MRTG which would allow a zero-based value, and if I had the 
time I might try and change that (although the new routines would need to 
work with the old data to show the history).  I don't like the idea of 
having to use two graphs, though - it would take up more space and look 
less elegant (I feel).

Yes, the one-day average you might expect to be near zero, but look at:

http://www.david-taylor.myby.co.uk/mrtg/hermes_ntp.html

where there is a transient in early January after a network problem at my 
ISP (routing change) and

http://www.david-taylor.myby.co.uk/mrtg/stamsund_ntp.html

where there is quite a lot of noise due to the heavy interactive use under 
Windows 2000.

Cheers,
David