Problem with Linux Machine's Request for Time from an XP Machine

Discussion in 'Linux Networking' started by W. Watson, Jan 9, 2005.

  1. W. Watson

    W. Watson Guest

    I'm awake again. Which of the three docs to you use for the ntp approach?
    Are your registry changes something you derived from one of these docs?

    See my prior post to the NG made about 4 hours ago. I'm about to read these

    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: <>
    W. Watson, Jan 11, 2005
    1. Advertisements

  2. W. Watson

    W. Watson Guest

    I believe you are misreading something. Also see my remarks below about 30
    or more lines.

    Like all work, scientific work requires judgement on available resources
    Time and money, for example. Once I get the information, I act accordingly
    in a way that applies to my situation. Frankly, if I followed every piece
    of advice I've gotten on this matter, I would be implementing about six
    different methods. This is *not* to say that I have not appreciated all the
    advice I have gotten. It has been helpful.

    If you will note another post below from snowbat at the end of this thread,
    I think he may have successfully broken through in an area that is
    appealing to me. Implementing a true NTP server on XP Pro. I'm verifying
    this at the moment.

    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: <>
    W. Watson, Jan 11, 2005
    1. Advertisements

  3. W. Watson

    Bill Unruh Guest

    From what I have read, MS ONLY impliments SNTP. It is also not at all clear
    that you have sufficient control of the on board clock so that you can slew
    it to compensate for hardware clock rate errors. This is critical to the
    operation of an NTP server.

    You want ms accuracy.
    a) You will NOT get that even with full ntp from a phone connection. Far
    too many latencies within the modems, etc. Thus if you want ms accuracy
    which you appear to need, a phone connection is NOT good enough.
    b) You will not get that with SNTP over the net.

    Scientific work requires that you do it right. If you do not have the
    resources to do it right then what is point of doing it. You will publish
    your results and people will be misled by them, since you will quote
    accuracies which are wrong.

    You need some way of getting ms (8 ms you state) accuracy. What is the
    value of the observations if you get 200ms accuracy?

    So it seems to me that you have three options.
    a) Get GPS receivers. This should give you microsecond accuracy. Whether
    the delays and latencies in your programs etc can use this kind of accuracy
    I do not know. You had better run the system as a real time system or
    program slicing is liable to destroy any timing accuracy.

    b)Get a real NTP server. I simply do not know what kind of accuracy you can
    get out of an NT box-- whether the OS has resources capable of delivering
    the accuracy you need even if you run NTp over a fast connection. Note
    that you will not get ms accuracy from even real ntp over a phone line
    modem connection. You will probably need at least DSL.

    c) abandon 8 ms accuracy and settle for 1 sec (well maybe 1/10 sec) accuracy.

    I think those three are your choices. For the last choice almost anything
    will probably work.
    It looks to me like what he is refering to is a SNTP server, not an NTP
    server. SNTP was developed for rough 1 second type accuracy for
    Bill Unruh, Jan 11, 2005
  4. W. Watson

    Snowbat Guest
    Primary guide for setting up.
    Information about the three W32Time service modes.
    Brief description of various time protocols including NTP/SNTP,
    NetRemoteTOD, and their own Domain Time products.

    Straight from the article.
    Snowbat, Jan 11, 2005
  5. I'm not guessing. "1 second a day" is a drift rate, which is
    That makes no difference.
    That makes no difference.
    That, as has been explained to you by others *in detail*, simply
    is not an option.
    I don't understand why you have to be told the same thing at
    least 5 times by at least 3 people before you listen. If it is
    a waste of time for me to comment, I won't...

    Read, again, what I wrote above.

    "Then both the XP box and the Linux box can safely access the
    Internet through a real firewall, and _both_ _can_ _also_ _run_ _ntp_
    _client_ _software_ _to_ _access_ _a_ _remote_ _server_..."

    However, given the accuracy you need, I'll agree with Bill and
    Christopher that you are on the edge as far as the accuracy that
    NTP over the Internet can possibly provide. I've been
    suggesting syncing to, but given your need for
    millisecond accuracy, that probably won't do it, especially
    through a modem connection.

    Regardless, the rest of the network topology suggest is still
    significant, and you can implement NTP as described, syncing
    to the nearest server available, as a temporary and imperfect
    solution, without any cost at all. Everything you do will
    be useful in a full fledged system that provides the accuracy
    you need.
    NTP over the Internet can barely provide you with that. You
    might get away with it, with a high speed connection, if you
    also try logging the time difference between your clock and say
    3 others, and then average the difference and use that as a
    correction factor for all of your logged observations.

    Regardless, it does appear that the only option which will
    provide sufficient accuracy is to install a local stratum 1
    clock source. You clearly do not have the technical background
    to construct one from component parts, which is probably the
    only way to accomplish it without significant expense.

    The problem area is the hardware interfacing of a frequency
    standard to a PC in a way that NTP can use. Obtaining a GPS
    unit with PPS output is only the initial step, and everything
    past that involves significant technical background. If you
    want to see what is involved, here is a very good description:

    Hence it appears that you need (either a good friend who can do
    these things, or) to purchase not just a frequency standard, but
    one that also provides a networked NTP server.

    Here are some examples. The first one looks very nice, and
    the price is $500. I didn't look for prices on others.
    Floyd L. Davidson, Jan 11, 2005
  6. He says he wants 8ms accuracy (and actually probably needs 1 ms). The above
    system, according to the information they provide at

    has the following specs:

    Masters and Independent Servers
    The default target accuracy is 250 milliseconds

    The default target accuracy is 100 milliseconds

    The default target accuracy is 500 milliseconds

    What good is that going to do for him?
    Floyd L. Davidson, Jan 11, 2005
  7. W. Watson

    W. Watson Guest

    Perhaps you should read the msgs from those who differ with you. Yes, NTP
    can be installed onto XP and yes the accuracy of 8ms can be met. Perhaps we
    are having a little confusion over ms, millisecond, that's about 1/10th of
    a second.

    I hope you don't mind if I prove the above to myself.

    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: <>
    W. Watson, Jan 11, 2005
  8. W. Watson

    W. Watson Guest

    I think I detect a misunderstanding in my stipulation of 8ms accuracy.
    Perhaps you folks think of that as 8 microseconds. To me it mean 8
    milliseconds. That is, a bit less that 1/10 th of a second.

    I think I can safely say this thread at its end (EOT). I have more than
    enough information to proceed in my own way. First, I plan to check out
    snowbat's very interesting discovery, and, if that fails, I will check out
    the claim that NTP of some sort better than what XP provides can be
    installed. If either of these fail, I have other goals to attain the
    accuracy needed, and will persue them as time and money allow. Presently,
    this discussion narrows down to snowbat and myself, privately.

    Thank you for your interest.
    W. Watson, Jan 11, 2005
  9. W. Watson

    prg Guest

    W.W., your w32time basic setup how-to is done and posted here. I had
    to wait for google to index it for a reliable link:

    I'll post the other sections there as well as they are of questionable
    interest in c.o.l.n except for your specific needs. Besides, it's less
    crowded -- I'm the only one who has responded to your post there ;-)
    see ya,
    email above disabled
    prg, Jan 11, 2005
  10. W. Watson

    W. Watson Guest

    Well, beore anyone becomes unglued over 8ms (millisecond) as nearly 1/10th
    of a second, let me say it is closer 1/100th of a second. Regardless, my
    partner and I will clear this up between us. Anyway, EOT.

    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: <>
    W. Watson, Jan 11, 2005
  11. Just take the time to actually read some of the material
    supplied, and tell us how you are going to get a system that
    won't keep better time than 500 ms to provide you with 8 ms
    Floyd L. Davidson, Jan 12, 2005
  12. W. Watson

    prg Guest

    Were you the other person I read today that confused 8 ms with 80 ms?
    8/1000=0.008 not 0.08 Yes, you are the same. You do get around.
    Good luck. But you're both wasting your time with the 3rd party
    software snowbat posted. It's not as capable as xp's w32time ;)

    You've been chasing a silver bullet solution all over 5-6 newsgroups
    for two months now and never staid with one long enough to see it
    through. Snowbat should take a look at your history on this project so
    he knows what he's in for. W.W. got almost all he needed for a
    solution in Nov. 04.

    You're both welcome to check the link in my last post -- or W.W. can
    just check his thread in microsoft.public.windowsxp.general

    You're even welcome to post for more help at that ng thread or read the
    other Section posts I plan to put there -- I may need them for future
    reference even if you do not. Once posted, Google grabs 'em, indexes
    'em, and makes 'em available with a browser -- a good way to archive my
    work ;-)

    Good luck,
    email above disabled
    prg, Jan 12, 2005
  13. W. Watson

    Bill Unruh Guest

    You are right. He has been running around in circles constantly for the
    past two or three months or more. And he never listens to anyone. Must get
    some sort of perverse pleasure out of riling up people who try to help him.

    Great. But he will probably start up a new thread and start all over again,
    hooking in some new poor suckers.

    Bill Unruh, Jan 12, 2005
  14. W. Watson

    Snowbat Guest


    Floyd, something strange is going on, either in your newsreader or mine.

    The URL I posted details eight time protocols:

    The URLs you quote me as having posted each have the first word of the
    following line wrapped on the end. This can also be seen in Google
    Groups. Specifically, the URL with the word 'Brief' wrapped
    on the end in your quote happens to give a sales pitch for Domain Time II:

    Your reply indicates you may have been reading the sales pitch? If that
    is the case, I refer you to the URL I actually posted that details eight
    time protocols. Otherwise, I fail to understand why have you singled out
    Domain Time II from the eight time protocols.

    Elsewhere in this thread, other posters considerably more qualified than I
    have discussed the suitability (or lack thereof) of SNTP for WW's project.
    I just wanted to address the problem listed in the subject - see if XP
    could serve time data in a format that a Linux box could sync to.
    Snowbat, Jan 12, 2005
  15. W. Watson

    Snowbat Guest

    What 3rd party software did I post? Are you having the same URL issue
    as "Floyd" ;-)
    Snowbat, Jan 12, 2005
  16. Sorry for not trimming that off. It is indeed a bug in the GNUS
    newsreader version that I'm using, where the word following a
    URL is not separated by whitespace. The results are what you've
    seen, and yes it is annoying!
    It appeared to be the one of choice.
    I have no problem with the fact that XP can indeed run a time
    server that Linux can sync to. That has never actually been in
    question by anyone other than the OP. The question for the rest
    of us is how accurate is it, and how flexible is it.

    The specs on the pages you reference are interesting, though it
    is extremely difficult to find meaningful data.

    For example,

    has a lot of interesting data. It does not mention accuracy,
    but does talk about "precision". Note that on that page it says
    the "precision" for Domain Time II (UDP & TCP) is "5-10
    milliseconds typical".

    Note that the description of "NTP/SNTP" discusses SNTP (rather
    than NTP), and apparently uses the terms precision and accuracy
    as synonyms. Whatever, it says the precision/accuracy is
    typically about 50-250 milliseconds.

    Whatever, clearly the only protocol that even comes close to
    being suitable for 8 ms accuracy is Domain Time II.

    The problem, for the OP, is that 5-10 ms *precision*, with a 500
    ms *accuracy* is not even close to accurate enough. And buried
    in the rest of their technical discussion the do say that the
    target accuracy (as opposed to the precision) is 500ms for the
    Floyd L. Davidson, Jan 12, 2005
  17. W. Watson

    prg Guest

    Actually, I have no idea what 3rd party software may have been eyed. I
    was just commenting on dr. Watson's "...I will check out the claim that
    NTP of some sort better than what XP provides can be installed", and
    "... this discussion narrows down to snowbat and myself, privately." I
    know W.W. ain't gonna install no stinkin' 3rd party wares without help
    -- presumably, you. Maybe not, if you're lucky.

    His connection is of such a quality that there is no such beast as
    "better than what XP provides". The connection quality -- rather the
    lack thereof -- will swamp any "better" (real and imagined) ntp

    As for urls, I did go to the protocols page. It was so laughable I
    looked no further into their offerings -- why would I?
    Youse guys have fun and good luck,
    email above disabled
    prg, Jan 12, 2005
  18. What "URL issue" did I have, other than reading farther than the
    sales pitch...
    Floyd L. Davidson, Jan 12, 2005
  19. W. Watson

    prg Guest

    When talk of GPS came up I wondered if you might comment. Imagined you
    had probably had more than average exposure to the use of such gizmos.
    Thanks for your input -- learned a thing or two...or three :)
    email above disabled
    prg, Jan 12, 2005
  20. Since you brought it up, let me just mention the fine print for the
    record. I use a GPS 35 to keep my machine at home in sync (side effect
    of some network measurement work). It seems to need to "see" a good
    chunk of the sky in order to work well. It took a bit of
    experimentation to get it placed just right. A house at the bottom of
    the north side of a hill might be problematic.
    Allen McIntosh, Jan 12, 2005
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.