[ntp:questions] Re: NTP clients not throttling back is this behaviour RFC compliant?

Wolfgang S. Rupprecht wolfgang+gnus20031203T040718 at dailyplanet.dontspam.wsrcc.com
Wed Dec 3 12:28:35 UTC 2003


> I've been thinking of building a board with FPGA and GPS card.
> The idea is that one read from the card would give you the
> correct time stamp.

I saw something like that on phk's old clock page.  It looked like a
neat idea if one could get the FPGA development board at a reasonable
price.  The one he used doesn't really qualify -- I think it was $3k
the last time I looked.

I don't know if you even need an FPGA for a counter.  The GPS has the
time to a few nS inside it.  That would be the place to get it from if
one could modify the firmware so that the GPS presented the high
resolution clock info to the outside world upon request.

Some companies like ublox.com did offer to give one the GPS firmware
under an NDA and agreement that one wouldn't ifdef out the tests for
60kft and 1000meters/sec. *

The other potential disadvantage of a clock in the FPGA is that the
clock probably couldn't be as fast as the cpu's cycle counter (and
probably couldn't be as high a resolution as the GPS's time either.

> I think we all agree that the main cost of processing a packet is
> kernel overhead.  Yes, you can save a lot of cycles if you send
> the packet back where it came from.  Are there interesting cases
> where that isn't good enough?

Well...  A network that had two routers where there was a
filter-enforced in-router and out-router.  I can't imagine anyone
would do that though and not expect some problems.

-wolfgang

* I believe this is a munitions export restriction, but if one didn't
export the GPS it is not clear why one would have to worry about this
nonsense in the first place.
-- 
Wolfgang S. Rupprecht 		     http://www.wsrcc.com/wolfgang/
       The above "From:" address is valid.  Don't mess with it.



More information about the questions mailing list