Ask Bjørn Hansen wrote:
> All their users are in a local network, so there's no good reason not 
> to run a local ntpd to use.
I fully agree with that!
Hopefully we sometime convince them to do that...
Until then, I think they are not the worst abusers one could imagine, 
and maybe a good test case.
>
> And besides that:  They are distributing an appliance.   The idea is 
> if you distribute something with a default configuration you need a 
> vendor pool.
That would be the idea, but I cannot blame them for not thinking about 
that even after reading the vendor FAQ.

Look, those people are not stupid.  They know about accurate time (most 
ISPs would not give a damn), they apparently have a method to 
pre-configure their routers or have their customers apply a standard 
configuration.
Those things require more thought and planning than the average ISP 
shows.  They may be unresponsive, but that could just be a language problem.
>
> Even the planned DNS improvements wouldn't help us here, because 
> presumably all these appliances (CPEs) are sharing a few DNS servers 
> so they'll keep banging on a small set of NTP servers for minutes at a 
> time (whatever we set the ttl to).
Well, it would at least reduce the length of the peaks.  Maybe to a 
duration where the average server operator no longer notices or cares.
Of course, the shorter the TTL the higher the load on the DNS servers 
from these folks.

Are you cooperating with Guillaume on the DNS improvements, or is his 
geo dns experiment an entirely separate project?

Rob
_______________________________________________
timekeepers mailing list
[email protected]
https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers

Reply via email to