Unfortunately, what we mean is that, even aggregated, this data could still be used to identify people.  This is because you could combine multiple data sources and infer a lot, especially from very active editors on smaller projects.

 

Actually, what Toby was describing above was at a country level.  Even that is too sensitive to do broadly.  But we are looking at figuring out how to publish aggregated data that is "aggregated enough".  Which means we have to determine at what point the different streams of information we have can not be used in combination to attack a person's identity.  As a side note, city geolocation is only 80% accurate from what I heard last, and it's especially inaccurate in places like India.  So if you do end up using anonymous IP geolocation you should take a deeper look into that.