Re: [gpsd-dev] Clarifications needed for the time-service HOWTO

[Harlan Stenn]

"Eric S. Raymond" writes:
> My plan is to actually run the whole tuning and measurement procedure
> we describe using the EVK 6H on RS232.  Maybe build some tools

NTF and the NTP Project would love to help with this.

> > Best way to play this game is not to play.  We can't agree on 'typical'
> > so just don't go there.  Just say that stratum is just a hop count.
> 
> Er...I don't think so.  For practical purposes that's the single most
> important error bound in the entire document.  If we say nothing about it
> the readers will feel cheated, and I believe justifiably so.
> 
> The proper thing to do here, I think, is give a range of measurements
> and say "This is what we've observed.  These are other peoples'
> observations. Take with small Siberian salt mine."

If there are tools for running these tests and measurements and folks
can do this themselves (perhaps automated) NTF will happily aggregate
and post the results.

> Even if we end up with an order-of-magnitude range and a note saying
> "For such an important number this one is controversial and hard to
> pin down" that will be better than punting the issue.
> 
> And now you know why I want to install Macx-1s on a bunch of routers
> and profile the hell out of NTP.
> 
> Harlan, what's the NTP project's figure for NTP time error bound?

Short answer, I don't know.  It really depends on a lot of things
between you and "the other guy".

Also see http://www.ripe.net/data-tools/stats/ttm .

If DLM wrote about this the answer is in the copy of The Book that I'll
be close to early next week.

H