Re: Weird dates and time stamps on old Linksys WAP11...

On Thu, 27 Jul 2006 10:18:14 -0700, Jeff Liebermann
<jeffl@xxxxxxxxxxxxxxxxxxxxxx> wrote in
<jathc2h2mnqu88m2qb3uutf40umrparu0v@xxxxxxx>:

John Navas <spamfilter0@xxxxxxxxxxxxxx> hath wroth:

I think it's a poor approach since it won't reliably result in a good
server, and since subsequent updates may come from quite different
servers, increasing time jitter and making clock drift correction less
workable. Better to carefully pick your own server(s).

Sure. A nearby time server and corrected clock makes more sense for
time critical applications. NTP clients all have smooth transition
features so there are no abrupt changes in time sync.

Not all by any means.

guess that's
necessary for the nifty Symmetricom client, which updates the time
once per hour. Were this client to use the NTP pool, then there would
certainly be some jitter. However, were the client to only update
once per day and when the machine starts, there would be no jitter and
sufficient accuracy for most home users.

While that would reduce frequency of correction (period), it would
almost certainly result in more, not less, jitter (amplitude).

While high time accuracy isn't needed much of the time for home users,
it can be important, sometimes very important, so, given that high
accuracy is so easy to achieve, I don't think it makes sense to
compromise. Kind of like not wearing a seat belt because the chance of
an accident is so low. ;)

I'm reminded of a client situation in which having the exact time
something had occurred was very helpful in working out an online issue
that otherwise could have been messy and/or expensive.

I usually use the NIST time sync software:
http://tf.nist.gov/service/its.htm
It's old, not very elegant, but works well enough.

I think it's actually pretty crappy -- in addition to some annoying
bugs/glitches, and abrupt clock changes, haven't you noticed how much
jitter it has even when querying Stratum 1 servers over lightly-loaded
broadband?

I wouldn't go much less than a day to avoid pounding on the designated
NTP server.

I think a good time client ought to be capable of picking an appropriate
interval itself based on observed clock drift.

--
Best regards, FAQ for Wireless Internet: <http://Wireless.wikia.com>
John Navas FAQ for Wi-Fi: <http://wireless.wikia.com/wiki/Wi-Fi>
Wi-Fi How To: <http://wireless.wikia.com/wiki/Wi-Fi_HowTo>
Fixes to Wi-Fi Problems: <http://wireless.wikia.com/wiki/Wi-Fi_Fixes>
.

Relevant Pages

  • Re: Weird dates and time stamps on old Linksys WAP11...
    ... server, and since subsequent updates may come from quite different ... features so there are no abrupt changes in time sync. ... Were this client to use the NTP pool, ... but don't really solve the fundamental problem of clock drift. ...
    (alt.internet.wireless)
  • Re: syncing two machines, microsecond precision?
    ... on moe I added the 'iburst dynamic' flags to the server ... client where I take the timestamp at the server, ... ntpq> rv ...
    (comp.protocols.time.ntp)
  • Re: Single GPS/PPS time source gets marked as a falseticker
    ... jitter. ... server 127.127.28.1 minpoll 4 maxpoll 4 prefer ... time correct to within 0 ms ... polling server every 16 s ...
    (comp.protocols.time.ntp)
  • Re: New to NTPD
    ... A 1 is shifted in from the right for each response from the server and a 0 for each failure to respond; 0 is bad 377 is good. ... The jitter, a measure of the random noise in the time received; ... Sunburn is a server on my LAN using a WWV receiver as reference clock; it's the middle of the day here and both signal reception and performance are extremely poor. ... Internet-A, Internet-C and Internet-D are part of the "selection" set and act as an "advisory committee". ...
    (comp.protocols.time.ntp)
  • [Fwd: Re: problems setting up IFF autokey]
    ... ind assID status conf reach auth condition last_event cnt ... server 0.pool.ntp.org ...
    (comp.protocols.time.ntp)