[atlas] Question about rt definition of DNS measurements

On Fri, Oct 04, 2013 at 11:50:11AM +0200,
 Klaus Darilion <klaus.mailinglists at pernau.at> wrote 
 a message of 14 lines which said:

> When doing DNS measurements I see that the average "rt" time is
> considerable higher when setting the DO bit.

I cannot reproduce it. Are you sure it isn't a caching effect? I tried
on a hot cache (using the same set of probes for three measurements,
one with the DO bit to warm the cache, then one without the DO bit and
one with, both using a hot cache). Compare public measurements
#1032220 and #1032221, the one with the DO bit is even a bit faster:

Without DO :
Measurement #1032220 for www.afnic.fr uses 488 probes
467 probes, 432 got a reply, 15 replies unparseable, 0 probes were apparently using a validator
Average RT is 66.67 ms
Test done at 2013-10-06T15:36:25Z

With DO :
Measurement #1032221 for www.afnic.fr uses 489 probes
478 probes, 424 got a reply, 18 replies unparseable, 144 probes were apparently using a validator
Average RT is 65.26 ms
Test done at 2013-10-06T15:40:57Z