Re: [gpsd-dev] PPS over USB

[Hal Murray]

address@hidden said:
> But we're not sure we can trust NTP.  That's the whole problem; if
> bufferbloat really is inducing very large, very short-period latency spikes,
> it may be screwing with the symmetry and statistical-smoothness assumptions
> that NTP synchronization relies on.

It's pretty obvious that ntpd gets confused if you have a long period of 
asymmetric routing.  I (and probably you) can easily generate that by 
downloading a CD or youtube show.  Probably the worst case is browsing the 
web with a browser that tries to be helpful by loading lots of pictures in 
parallel and/or preloading stuff that you might click, or doing that while 
you have a download going.

My DSL connection sees occasional queuing delays of over 3.5 seconds.

It might be possible to avoid most of the problems with a big enough huffpuff 
parameter and/or appropriate tweaks to other parameters.  I haven't studied 
this area carefully.


> One of the explicit goals of the Cosmic Backround Bufferbloat Detector is to
> sanity-check NTP. 

If you see anything that seems unreasonable, please let me know.  ntpd has a 
huge collection of heuristics (hacks?) that try to filter out all sorts of 
quirks/glitches.  I'm sure there are some cases that will get past the 
current filters.



-- 
These are my opinions.  I hate spam.