On Saturday 26 November 2005 16.22, Nelson Minar wrote:
> Another way to measure clients is to count DNS requests to pool.ntp.org
>
> > the best idea would probably be to ask operators to volunteer what data
> > they already are collecting
>
> The DNS load balancing is pretty much fair, right? Just biased to
> goodtickers? If so, extrapolating from a single server's logs should get
> us close.

No, the DNS round robin is pretty much randomly unfair.  I think it might be 
more or less reliable if you average over long enough time periods (a 
week?), but for anything shorter, caching DNS servers (partly with bogus 
expiry algorithms etc.) will probably distort the data quite a lot.

But getting a very rough base estimate should be possible, with a +/-100% 
error margin or so.

-- vbi

-- 
Windows: the ultimate triumph of marketing over technology.

Attachment: pgpyT0mxZqTf1.pgp
Description: PGP signature

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

Reply via email to