Getting the Real NTP Server Active on XP Pro
W. Watson
on XP Pro. Some urged me to use NTP instead. I finally began trying this
approach and late Dec. and have not had a lot of luck. Here's what I would
like to do. I have two computers, a Linux box and XP box, on a local
network. I'd like to set the clock on the XP box daily from an NTP source
on the internet. That's been easy to do. At some point during the day, I'd
like the linux box to request from XP the correct time. That has not been
easy at all. I've been told that XP does not really have NTP, but does have
sntp. Any request to XP (on 192.168.0.3) goes unheeded.
People in another group (a linux networking NG) have suggested that I need
to replace sntp with ntp. This most likely requires a download. I haven't
googled around for it yet. There's a easily found document on the web
titled something like an autoritative time source for XP that *seems* to
suggest a registry change will create an NTP server on XP. So far following
the instructions there has led to an impass. This morning I received a
message from a fellow on the linux network NG saying he had found yet
another registry change doc, a simpler one, on the net that created an NTP
server in XP for him, and he was able to make a request from a linux box to
it for time.
So what in the last paragraph is true? 1. That is, is the only way to get a
good ntp server on xp to download some code? 2. Does the authoritative doc
really work? 3. Is a simple registration change an easy way to get an ntp
server on XP? Any experience with any of these three items would be
appreciated.
BTW, I'm looking for 8ms of accuracy to the linux box. Yes, I know using a
linux server instead of the current config of ntp for XP is a better way to
do this.
Here are the two sources for the registry change approach:
<http://support.microsoft.com/default.aspx?scid=kb;en-us;314054>
<http://www.windowsnetworking.com/articles_tutorials/Configuring-Windows-Time-Service.html>
Two other related articles:
<http://dtdocs.ntp-systems.com/kb/kb2001.002.asp>
<http://www.greyware.com/software/domaintime/technical/network/protocols.asp>
--
Wayne T. Watson (Watson Adventures, Prop., Nevada City, CA)
(121.015 Deg. W, 39.262 Deg. N) GMT-8 hr std. time)
Obz Site: 39° 15' 7" N, 121° 2' 32" W, 2700 feet
Web Page: <home.earthlink.net/~mtnviews>
ABC
"W. Watson" <wolf_...@invalid.inv> wrote in message
news:pRSEd.4219$pZ4...@newsread1.news.pas.earthlink.net...
Personally, I would use the Linux box to get the correct time from one of
the public ntp servers in the internet. Linux has a built-in ntp server,
which you could get the XP box to check the time and correct its clock.
You can use a prog called Automachron from
http://www.oneguycoding.com/automachron/.
S
Darren Dunham
> Back in mid I asked this group about using my Garmin 12XL as a time source
> on XP Pro. Some urged me to use NTP instead. I finally began trying this
> approach and late Dec. and have not had a lot of luck. Here's what I would
> like to do. I have two computers, a Linux box and XP box, on a local
> network. I'd like to set the clock on the XP box daily from an NTP source
> on the internet. That's been easy to do. At some point during the day, I'd
> like the linux box to request from XP the correct time. That has not been
> easy at all. I've been told that XP does not really have NTP, but does have
> sntp.
Perhaps it has an SNTP client, but you can always install the full NTP
packages.
While normally I'd set up the Linux box as the timeserver, you can do it
this way. I'd just expect somewhat less accuracy.
> People in another group (a linux networking NG) have suggested that I need
> to replace sntp with ntp. This most likely requires a download. I haven't
> googled around for it yet.
www.ntp.org and the FAQ there is a good place to start.
There's a easily found document on the web
> titled something like an autoritative time source for XP that *seems* to
> suggest a registry change will create an NTP server on XP. So far following
> the instructions there has led to an impass. This morning I received a
> message from a fellow on the linux network NG saying he had found yet
> another registry change doc, a simpler one, on the net that created an NTP
> server in XP for him, and he was able to make a request from a linux box to
> it for time.
> So what in the last paragraph is true? 1. That is, is the only way to get a
> good ntp server on xp to download some code? 2. Does the authoritative doc
> really work? 3. Is a simple registration change an easy way to get an ntp
> server on XP? Any experience with any of these three items would be
> appreciated.
There may be several bugs in the windows server. Try looking at some of
the FAQ entries for it...
http://www.ntp.org/ntpfaq/NTP-s-trouble.htm#S-TRBL-SPEC-WINNT
Also, there's a lot of folks running this (including on windows) on
comp.protocols.time.ntp.
While windows may not be the best NTP platform, if you don't have PPS
support and you're just running NMEA from a GPSR, that would probably be
the biggest source of error.
--
Darren Dunham ddu...@taos.com
Senior Technical Consultant TAOS http://www.taos.com/
Got some Dr Pepper? San Francisco, CA bay area
< This line left intentionally blank to confuse you. >
Frank Looper
news:pRSEd.4219$pZ4...@newsread1.news.pas.earthlink.net...
> on XP Pro. Some urged me to use NTP instead. I finally began trying this
> approach and late Dec. and have not had a lot of luck. Here's what I would
> like to do. I have two computers, a Linux box and XP box, on a local
> network. I'd like to set the clock on the XP box daily from an NTP source
> on the internet. That's been easy to do. At some point during the day, I'd
> like the linux box to request from XP the correct time. That has not been
> easy at all. I've been told that XP does not really have NTP, but does
have
> sntp. Any request to XP (on 192.168.0.3) goes unheeded.
>
8 msec is extremely doubtful, but the clock in your computer will not hold
that on even an hourly update, much less a daily one. It also will work as a
timeserver on your WinBox, and can be set to update at any frequency, in
seconds.
Frank
Bjorn Gabrielsson
> People in another group (a linux networking NG) have suggested that I
> need to replace sntp with ntp.
Are you sure it was not comp.protocols.time.ntp? Kind of remember you...
People there are not linux centric. Most of the ntp gurus prefer
FreeBSD. You were suggested ntp because that implements the
whole protcol - why not use the _real thing_ when its available for
free and with good support from the programmers.
> This most likely requires a download.
Yes! Is that a problem?
> I haven't googled around for it yet.
You do not need to google for it. I think you got the link already at
'the other' NG. But here it is again.
Good luck!
W. Watson
Can't do that.
W. Watson
I'm afraid XP will have to do. Linux app stands in the way. Thanks for the
references.
W. Watson
8ms = 0.08 sec or about 1/10 of a second.
W. Watson
> "W. Watson" <wolf_...@invalid.inv> writes:
>
>
>>People in another group (a linux networking NG) have suggested that I
>>need to replace sntp with ntp.
>
>
> Are you sure it was not comp.protocols.time.ntp? Kind of remember you...
> People there are not linux centric. Most of the ntp gurus prefer
> FreeBSD. You were suggested ntp because that implements the
> whole protcol - why not use the _real thing_ when its available for
> free and with good support from the programmers.
Yes, that is one of the places I probed, and got my knuckles light rapped
for mentionining XP. :-)
>
>
>>This most likely requires a download.
>
>
> Yes! Is that a problem?
Only in sofar as I need to know where to find the download. I think another
poster pointed to it. Ah, I see your remark below.
>
>
>>I haven't googled around for it yet.
>
>
> You do not need to google for it. I think you got the link already at
> 'the other' NG. But here it is again.
>
> http://norloff.org/ntp/
>
> Good luck!
Stan Gosnell
> 8ms = 0.08 sec or about 1/10 of a second.
No, 8ms = .008 sec. 80ms = .08 sec. A millisecond is a thousandth of a
second, not a hundredth.
--
Regards,
Stan
Frank Looper
>
seconds, or did I just read the original number wrong? If you just need to
within 1/10 second, the app I mentioned will work great. Your best bet,
however, will be to slave to a GPS that can output 1pps. However, the
original post said only a GPS 12XL was available, so that can't happen, I
don't think. I have a GPS 12, and it doesn't output 1pps, but maybe the XL
does.
Frank
Terje Mathisen
> W. Watson <wolf_...@invalid.inv> wrote:
>
>>Back in mid I asked this group about using my Garmin 12XL as a time source
>>on XP Pro. Some urged me to use NTP instead. I finally began trying this
>>approach and late Dec. and have not had a lot of luck. Here's what I would
>>like to do. I have two computers, a Linux box and XP box, on a local
>>network. I'd like to set the clock on the XP box daily from an NTP source
>>on the internet. That's been easy to do. At some point during the day, I'd
>>like the linux box to request from XP the correct time. That has not been
>>easy at all. I've been told that XP does not really have NTP, but does have
>>sntp.
> There may be several bugs in the windows server. Try looking at some of
> the FAQ entries for it...
Indeed. Windows comes with a program that claims to be ntp, but in
reality is a broken implementation of sntp (Simple NTP). :-(
> http://www.ntp.org/ntpfaq/NTP-s-trouble.htm#S-TRBL-SPEC-WINNT
>
> Also, there's a lot of folks running this (including on windows) on
> comp.protocols.time.ntp.
Right, that's a much better location (actually the only good one) for
this query. I maintain the windows binaries for the official ntpd
distribution, and I can only second all the suggestions to use the *ix
machine as the ntp server!
>
> While windows may not be the best NTP platform, if you don't have PPS
> support and you're just running NMEA from a GPSR, that would probably be
> the biggest source of error.
The first/main problem here is that most regular consumer GPSs without a
PPS (Pulse Per Second) output has a really, really horrible jitter in
the serial port timing output.
+/- 500 ms is not too unusual!
It will still work though, after sufficient amounts of filtering.
Terje
--
- <Terje.M...@hda.hydro.com>
"almost all programming can be viewed as an exercise in caching"
Terje Mathisen
> 8ms = 0.08 sec or about 1/10 of a second.
No way.
8 ms = 0.008 sec or less than 1/100 of a second.
If your real requirement is for 80 ms, then the chances of getting there
with a non-timing GPS and windows are better.
Terje Mathisen
> Darren Dunham wrote:
>> While windows may not be the best NTP platform, if you don't have PPS
>> support and you're just running NMEA from a GPSR, that would probably be
>> the biggest source of error.
>>
> I'm afraid XP will have to do. Linux app stands in the way. Thanks for
> the references.
'Linux app stands in the way'???
The only possible way this can be right would be if you have some Linux
app that requires the NTP port (port 123), but doesn't implement NTP
itself! Is this correct?
Joop van der Velden
> The first/main problem here is that most regular consumer GPSs without a
> PPS (Pulse Per Second) output has a really, really horrible jitter in
> the serial port timing output.
>
> +/- 500 ms is not too unusual!
>
> It will still work though, after sufficient amounts of filtering.
Then again - NMEA was never meant or specified for accurate timing
purposes. Better use the native binary mode, in combination with PPS.
--
Joop van der Velden
pe1...@amsat.org
do...@xrexxgetti.usenet.us.com
> The only possible way this can be right would be if you have some Linux
> app that requires the NTP port (port 123), but doesn't implement NTP
> itself! Is this correct?
Or some serial device is plugged in to his only serial port.
--
---
Clarence A Dold - Hidden Valley (Lake County) CA USA 38.8-122.5
Terje Mathisen
I've seen results indicating that Garmin binary can give you less
jitter, but that still doesn't provide the missing PPS signal.
As long as you do have PPS, even NMEA is enough to number the seconds.
Terje Mathisen
> Terje Mathisen <terje.m...@hda.hydro.com> wrote:
>
>>The only possible way this can be right would be if you have some Linux
>>app that requires the NTP port (port 123), but doesn't implement NTP
>>itself! Is this correct?
>
> Or some serial device is plugged in to his only serial port.
That would be easy to fix: With the NMEA jitter, adding either an extra
SIO card, or an external Serial-2-USB converter would still be OK.
ABC
"W. Watson" <wolf_...@invalid.inv> wrote in message
You say you can't do that,. but haven't said why?
S
do...@xrexxgetti.usenet.us.com
> do...@XReXXGetti.usenet.us.com wrote:
>> Or some serial device is plugged in to his only serial port.
> That would be easy to fix: With the NMEA jitter, adding either an extra
> SIO card, or an external Serial-2-USB converter would still be OK.
A second serial port might not be possible due to lack of slots. USB might
not exist. If it does, I wasn't sure if USB-Serial was as simple in Linux
as in Windows, and I haven't even tried it there, so I didn't suggest it.
Of course, we don't know, and this is the second thread on the same topic,
ending with "I can't try anything that I don't already have working", with
no other details.
Bjorn Gabrielsson
> Joop van der Velden wrote:
>
> > Terje Mathisen wrote:
> >
> >> The first/main problem here is that most regular consumer GPSs
> >> without a PPS (Pulse Per Second) output has a really, really
> >> horrible jitter in the serial port timing output.
> >>
> >> +/- 500 ms is not too unusual!
> >>
> >>It will still work though, after sufficient amounts of filtering.
> > Then again - NMEA was never meant or specified for accurate timing
> > purposes. Better use the native binary mode, in combination with PPS.
>
> I've seen results indicating that Garmin binary can give you less
> jitter, but that still doesn't provide the missing PPS signal.
>
> As long as you do have PPS, even NMEA is enough to number the seconds.
>
Joop is used to Jupiter receivers. Follow his advise when dealing with
Jupiters. They (the one I have) have the little feature of beeing either
some 1.02s _or_ 2.02s late with NMEA messages. They _might_ stay in
either state for long times. Jitter at each state is not bad, a few
ms. But not knowing if it is one or two seconds late is really bad!
It is kind of sneaky too. You need to run the receiver for hundreds of
hours to be sure to see it.
/Björn
Joop van der Velden
> > > Then again - NMEA was never meant or specified for accurate timing
> > > purposes. Better use the native binary mode, in combination with PPS.
> >
> > I've seen results indicating that Garmin binary can give you less
> > jitter, but that still doesn't provide the missing PPS signal.
> >
> > As long as you do have PPS, even NMEA is enough to number the seconds.
> Joop is used to Jupiter receivers. Follow his advise when dealing with
> Jupiters. They (the one I have) have the little feature of beeing either
> some 1.02s _or_ 2.02s late with NMEA messages. They _might_ stay in
> either state for long times. Jitter at each state is not bad, a few
> ms. But not knowing if it is one or two seconds late is really bad!
>
> It is kind of sneaky too. You need to run the receiver for hundreds of
> hours to be sure to see it.
Hi Bjorn - good to see you again,
For both Linux and Windows platforms there are clock drivers that use
the Rockwell Jupiter's binary mode, so that should solve this problem.
My DIY gps project is frozen for the moment, but i'm still looking for
a followup. There is a guy at ebay selling Jupiters, a bit more recent
than the ones i had so far. It's a 3.3V version. I got one from him
and still need to test it for this pecularity that you reported.
I also have an option for a Sirf-1 based module but altough is is from
the same time period as the Jupiters (1999) i get the feeling that the
sirf-1 is *way below* the performance of the Jupiters.
W. Wat son
Ah, you don't believe me. :-) No possibility of an internet connection or
local network location. The box is very isolated. GPS.
--
Wayne T. Watson (Watson Adventures, Prop., Nevada City, CA)
(121.015 Deg. W, 39.262 Deg. N) GMT-8 hr std. time)
Obz Site: 39° 15' 7" N, 121° 2' 32" W, 2700 feet
Most vehicle/deer accidents occur at sunset.
Vehicle deer whistles are ineffective.
Web Page: <home.earthlink.net/~mtnviews>
W. Wat son
I've been away. Let me see if I can recover from my absence. I see this
thread carried on without me.
Simply put the amount of effort and work to get Linux and XP to talk to one
another has already been daunting, and, more importantly, as someone
pointed out, I'm running up against NTP (Linux) vs XP (SNTP).
I'd really like to concentrate on GPS. I've looked at NTP/SNTP, hwclock,
and Chrony. At this point, amazingly, hwclock has given me the most
satisfaction. It may not be deadly accurate but I find it dependable in the
sense the drift is not unreasonable, and, very importantly, by using a cron
script in Linux, I can recover time accuracy pretty well after a system
crash. Chrony is probably a better choice, but I've run into Linux problems
and a problem just communicating with the Chrony subscription list and
getting any response from the Chrony author. Maybe they've gone on an
extended vacation? I'm about worn out on this stuff, but if I want to push
for better accuracy, then I think GPS is the choice.
Regarding GPS, it looks like the cheap way to do this is with a Motorola
Oncore (older version that's supposedly available for about $20-30), or a
Garmin GPS 25 (engine) for about $75 (they are discontinued and on sale). A
power supply, and maybe an antenna will be needed. I'll probably need to
putter around with the apps Linux code to make this work. For the mean
time, I'm backing down on spending a lot of time on this stuff. hwclock
looks like it's sufficient for present needs.
--
Wayne T. Watson (Watson Adventures, Prop., Nevada City, CA)
(121.015 Deg. W, 39.262 Deg. N) GMT-8 hr std. time)
Obz Site: 39° 15' 7" N, 121° 2' 32" W, 2700 feet
Most vehicle/deer accidents occur at sunset.
Jack Erbes
<snip>
> I've been away. Let me see if I can recover from my absence. I see this
> thread carried on without me.
>
> Simply put the amount of effort and work to get Linux and XP to talk to
> one another has already been daunting, and, more importantly, as someone
> pointed out, I'm running up against NTP (Linux) vs XP (SNTP).
NetTime is a simple utility that can be installed to run as a service on
95/98/NT/2000/XP machines. It may run on XP also. More info on it here:
http://sourceforge.net/projects/nettime
and here:
http://nettime.sourceforge.net/
I have run that for several years on w2k machines, I recently changed
one of my two w2k machines to use the Windows Time service and have not
seen any noticeable time keeping errors on either of the two machines.
For awhile I was frequently comparing the times on the machines to
http://nist.time.gov/timezone.cgi?Eastern/d/-5/java but found it
reliable and stopped doing it.
It may be worth trying NetTime to simply test the communications between
that, running as a service, and your linux machine. The developers or
development of NetTime seem to have some ties to the Unix community,
maybe that bodes well for getting it to work with linux.
Jack
--
Jack Erbes in Ellsworth, Maine, USA - jackerbes at adelphia dot net
(also receiving email at jacker at midmaine.com)