My only thought is that "city" makes me uncomfortable. Did we track
down a precise use case for that in the end?
Yes, the Los Alamos National Lab folks' proposal:
https://meta.wikimedia.org/wiki/Research:Geo-aggregation_of_Wikipedia_pagev…
We talked to them yesterday and it seems the time granularity is not as
important. That's why that dataset is *daily* and the other one is
*hourly*. Either way, these will be k-anonymized at any level. Once we
have some data up, though, I'd love for people who are good at this to try
and attack the datasets in combination and from different points of view
like t-closeness, etc. I don't want to leak any info and any help on that
is appreciated 'cause it's a hard problem.