[geogriffin]

Am I missing something, or is the reason most of the queries observed have low TTL because, well, they have a low TTL? IOW, the higher TTL responses would be cached downstream and so you'd see them less often. If that is the case, the distribution shown is not all that surprising.

[necovek]

It's weird how people are not understanding this: perhaps it's the way you phrased it. Or perhaps you missed to mention the core part from the article: the experiment was only run "for a few hours". This means that many a DNS record (well, most) with TTL greater than the experiment duration would not show up in the data.

FWIW, I've learned in the past that while there are plenty of people who claim to want communication to be as succint as possible, majority are unable to understand when somebody is really terse (while still saying exactly enough). I've learned to follow up such a terse statement with examples and longer explanations for the majority that does not get it.

But maybe it's just that people don't expect the mathematics-level precision on the internet :)