Path: utzoo!utgpu!watserv1!watmath!att!emory!gatech!usenet.ins.cwru.edu!agate!ucbvax!udel.edu!Mills From: Mills@udel.edu Newsgroups: comp.protocols.time.ntp Subject: Re: wwvb.isi.edu 15 minutes off? Message-ID: <9102241647.aa29177@huey.udel.edu> Date: 24 Feb 91 21:47:22 GMT Sender: daemon@ucbvax.BERKELEY.EDU Distribution: inet Organization: The Internet Lines: 33 Tim, On closer inspection at wwvb.isi.edu earlier today I found the Spectracom clock wandering 25 minutes slow, but ticking correct format and quality indicator. As of now (2129Z 24 Feb) it has returned to truetick. This is the first time I have seen this kind of wobble. All previous wobbles known to me have involved onboard clock silicon failures such as stuck hot and cold bits and microprocessor hiccups. What worries me is that the wwvb hardware has recently been gutted and replaced with the innards of the new Spectracom Netclock/II, which they just announced. I sure hope this particular wobble is not species-specific, since I would like to upgrade four of the other Spectracom 8170 clocks now used at other primary servers. The new clock, as well as the upgrade, now decode the year and leap information, which would be a Good Thing and also simplify my life in the first moments of every New Year. As Tim noted, several secondary sites got warped along with the wwvb server itself. Some ten out of about 120 chimers at wwvb warped tick in this way. This shouldn't happen with strategically configured peers as recommended in the pub/ntp/ntpinfo file on louie.udel.edu and in many messages sent to this list and published in various RFCs and other documents. In addition, I can say that a version-3 server has considerably more resistance to falsetickers than the previous versions now indigent to the present Unix daemons. In reviewing the statistics collected by a version-3 fuzzball near here during a five-day interval including the wwvb warp, I found the mean error 0.768 ms, standard deviation 1.782 ms and max error +4.922 -2.579 referenced to its other stratum-1 peers, including wwvb. However, during this interval 4 of the 9 configured peers became warped for at least long enough to get noticed. I suspect some of these warps were due to reboots and zoon, but it also reveals that the wwvb warp was gloriously ignored. Dave Brought to you by Super Global Mega Corp .com