Hacker News Reader

I agree that a balance can be struck, and a lot can be learned without sacrificing substantial privacy. I don't think it's the best of both worlds, though; some sacrifices must be made somewhere. Do you think the amount of information described in my prior comment (cursor movement records that can be filtered based on any demographic data from the profile that generated it) can be analysed at arbitrary points in the future without sacrificing some amount of privacy? Note that I'm not placing a value on this and saying such data should be collected, I'm just arguing that there is a tradeoff. Differential privacy systems do make such tradeoffs.

I think much of the value can be gained by collecting the data with some kind of differential privacy noise added to keep it from being traced back to any given user. But obviously data collected in this way is less useful--it may not matter, given diminishing returns, for services with enough scale, but yeah these techniques can make things harder for smaller services.

I don't know what to learn from these ideas.

This user speaks of Utopia where there is infinite engineering resources and endless profit to dig into.

Maybe I suck at business, but making profit is very hard.

The end result is the same though. Any client side tracking, privacy-friendly or not will be blocked. So many sites just use GA anyway - because some users will block regardless.