[ntp:hackers] Comments invited

David L. Mills mills at udel.edu
Sun Dec 28 22:12:45 PST 2003


P-H,

The monotonic-increasing behavior coincides with the Lamport
happens-before partial-order assertion. It's nice to be theoretically
pristine. As most of my old cesia have bit the dust, I am reduced to
GPS.

Dave

Dave

Poul-Henning Kamp wrote:
> 
> In message <3FEF1A4C.6BC3CD86 at udel.edu>, "David L. Mills" writes:
> 
> >During a leap second when
> >the clock is for all purposes frozen, an idiot in a reading loop could
> >torque the clock 2.5 microseconds into the first second after the leap.
> >Not Earth shaking for sure, but maybe Earth wiggling.
> 
> I pointed the deficiencies of that aproach out to you a long time ago.
> 
> That is why FreeBSD does not guarantee timestamps to be increasing,
> only to not decrease.
> 
> >I don't know if there are others than you with ambitious oscillators; we
> >ran those here for some years and I got tired of upgrading the interface
> >every time a new bus spec came along. First the PDP11 U bus, then the
> >LSI-11 Q bus, then the SPARC S bus, then the PSA bus, then the PCI bus,
> >then...
> 
> If you have a student with a soldering iron to spare, the soekris
> net4501 (www.soekris.com) is an eminent and cheap hardware platform
> for timestamping (http://phk.freebsd.dk/soekris-pps)  Just remove
> the stock xtal and feed it a 30-40MHz frequency from an atomic driven
> PLL chip.
> 
> --
> Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
> phk at FreeBSD.ORG         | TCP/IP since RFC 956
> FreeBSD committer       | BSD since 4.3-tahoe
> Never attribute to malice what can adequately be explained by incompetence.



More information about the hackers mailing list