Re: simple time server - NTP

This is a discussion on Re: simple time server - NTP ; >It takes a couple of hours for the program to figure out "what the time >is precisely". Remember that in this thread, I'm talking about a situation where the local clock of one computer is the time standard. If the ...

+ Reply to Thread
Results 1 to 2 of 2

Thread: Re: simple time server

  1. Re: simple time server

    >It takes a couple of hours for the program to figure out "what the time
    >is precisely".


    Remember that in this thread, I'm talking about a situation where the
    local clock of one computer is the time standard. If the time server
    is running on that computer, it knows by definition at all times
    precisely what time it is.

    There is a whole class of applications of network time service where
    it's important that a bunch of computers adhere to a time standard,
    but that standard doesn't have to be UTC. There's also a whole class
    of applications where wristwatch level precision is adequate. In my
    case, I can't imagine making use of more than about 100 ms precision.

    I think I've discovered in my brief investigation of the Ntpd
    reference implementation that it is not appropriate for those
    applications. I'm still trying to figure out whether NTP itself is.

    --
    Bryan Henderson Phone 408-621-2000
    San Jose, California
    _______________________________________________
    questions mailing list
    questions@lists.ntp.isc.org
    https://lists.ntp.isc.org/mailman/listinfo/questions


  2. Re: simple time server

    Bryan Henderson wrote:

    >>It takes a couple of hours for the program to figure out "what the time
    >>is precisely".

    >
    >
    > Remember that in this thread, I'm talking about a situation where the
    > local clock of one computer is the time standard. If the time server
    > is running on that computer, it knows by definition at all times
    > precisely what time it is.
    >
    > There is a whole class of applications of network time service where
    > it's important that a bunch of computers adhere to a time standard,
    > but that standard doesn't have to be UTC. There's also a whole class
    > of applications where wristwatch level precision is adequate. In my
    > case, I can't imagine making use of more than about 100 ms precision.
    >
    > I think I've discovered in my brief investigation of the Ntpd
    > reference implementation that it is not appropriate for those
    > applications. I'm still trying to figure out whether NTP itself is.
    >


    If you are willing to settle for time to the nearest second you can use
    rdate. If you need better; e.g. time to the nearest 100 milliseconds or
    better, SNTP *might* do it for you but I wouldn't want to rely on it myself.

    You can, if you wish, use ntpdate to set your clocks from some NTP
    server. You could run it in a cron job once an hour or something like
    that and it should keep your clock within a hundred milliseconds of the
    server unless you have a machine with a really bad local clock. Ntpdate
    does not provide the full functionality of ntpd and it is "deprecated"
    and sooner or later will be dropped from the distribution.

+ Reply to Thread