wbh at conducive.org
Wed Apr 13 04:07:36 PDT 2005
Joerg Sonnenberger wrote:
On Tue, Apr 12, 2005 at 10:17:48AM -0700, Matthew Dillon wrote:
I'm not even sure how well it works any more, but I do know that
the REST of the ntp code in the kernel works because I went through
and fixed it last year. It *is* complicated, even with the extra
comments I added.
My biggest problem is that the code does a lot in the kernel what
should be done in userland IMO.
Agree that. Better yet - in a 'hardware' plug-in' where the only latency
would be the I/O port & bus.
In anycase, we want to keep the pps stuff, it's something we need to
OK, I'll leave the ntp_adjtime related stuff untouched until my own work
is ready and I have thought how to best handle PPS.
Someone willing to donate a Caesium watch or a long wave receiver? :)
Have the second such not in critical use. Citizen WWVL wall-clock.
But Continental US (Eastern US at that..) viable only. Not expected
to work where you are ;-)
Also have several oven-controlled crystals, military and telco surplus.
Best of them IIRC, 1 part in ten million, worst 10 ppm.
Need -48 VDC for one type (octal vacuum-tube socket),
+/_12VDC or +5VDC for the other (PCB mount, about half the
size of a packet of cigarettes).
Also - If I haven't binned it - a 'D-Clock' FDC cable-attach with
software for CP/M and/or PCDOS. Co-exists with 'real' FDD.
These date from the '60's, '70's, maybe early '80's, IIRC.
State of the art at the time... But that was a long time ago...
Cheapest current-practice should be a GPS clock-only (no nav circuitry),
which needs only two satellites in 'view'.
Time-keeping over UDP/TCP is like the dancing Bear.
The marvel of the thing isn't that it is done well, but that it
can be done at all.
SONET & SDH are another animal entirely, and
one *can* extract a Stratum 3 clock (de-facto, not de-jure) from
a telco ADSL/SDSL, or ISDN NTE (but not, AFAIK, a cable modem).
More information about the Users