[ntp:questions] 500ppm - is it too small?

nemo_outis abc at xyz.com
Mon Aug 17 23:34:27 UTC 2009


Harlan Stenn <stenn at ntp.org> wrote in
news:ywn9r5varydm.fsf at ntp1.isc.org: 

>>>> In article <Xns9C6A7CAD22A5Cpqwertyu at 69.16.185.247>, "nemo_outis"
>>>> <abc at xyz.com> writes: 
...
> nemo> The justification for 500 ppm in "Das Buch" is very thin gruel
> and nemo> amounts to litle more than "argument by repeated assertion."
> 
> Could be, and I suspect (but do not know for sure) that there is more
> detailed information in some of the white papers at DLM's website at
> UDel. 
> 
> I'm sure some folks would appreciate whoever could either debunk these
> claims or clearly/cleanly show why they are correct, and also what
> would need to be done if one were to change this value.

I am not saying that 500 ppm "cannot" be justified, but rather that it 
*has not* been justified.  In the absence of such justification it stands 
accused (at least in my mind) of being arbitrary (particularly given the 
incomplete but tantalizing evidence to the contrary from Chrony).  

If someone can point me to solid evidence justifying 500 ppm I would be 
most grateful.  I am always delighted to "lose" an argument for that 
means I have increased my store of knowledge and been enriched thereby 
:-)

However, from the difficulties even adepts are having in laying hands on 
the justifying evidence, I believe I can say that, if nothing else, the 
documentation of the architecture and design decisions underlying ntp is 
- ahem! - somewhat obscure and inaccessible, at least on this point.

...
> nemo>  In fact, the book itself discusses a 1997 survey (section 6.6)
> that nemo> shows time offsets of hundreds of milliseconds, so far from
> being nemo> unusual, are common. This (inter alia) makes the book's
> light regard nemo> for large offsets seem more than a little cavalier.
> 
> You are missing the point.  it's not about *time* offsets, it's about
> *frequency* offsets.

I take what little evidence there is that appears to bear on the issue 
and use it as best I can.  If you have better, please share.
 
...
> In my experience Prof. Mills is (sometimes eventually) open to new
> ideas, but since he is so *thoroughly* steeped (both in depth and in
> breadth) in this arena and since he has been down so many of these
> roads with so many people so many times before, his usual response is
> along the lines of "Been there, done that, wore out the t-shirt, but
> if you really think I'm wrong please code it up, run it thru rigorous
> and encompassing analysis and simulation, and if at the end of all of
> that you still have something we can talk."

It is possible, even likely, that we are rewalking a well-trodden path.  
Well-trodden, perhaps, but not well-lit.

It is not a question, except peripherally, how well versed Prof Mills is 
with the issues and how knowledgable he is about the history, including 
unproductive dead ends - *unless* that information is readily available 
to others to make an informed judgment about whether and how to use ntp, 
its internal mechanisms, and its limits.  In short, I'm not looking for 
unchallengeable "ex cathedra" pronouncements from a Pope (no matter how 
well informed he is and no matter his ennui with tiresome questions from 
acolytes).  No, I'm looking for info I can evaluate myself. 

Regards,




More information about the questions mailing list