4 ms·
It's described here, how they deal with that sort of stuff: https://gist.github.com/solso/423a1104a9e3c1e3b8d7c9ca14e885e5 https://gist.github.com/solso/423a110
by 482794793792894 9y ago
It's described here, how they deal with that sort of stuff: https://gist.github.com/solso/423a1104a9e3c1e3b8d7c9ca14e885e5 https://gist.github.com/solso/423a1104a9e3c1e3b8d7c9ca14e885...
- Vinnl 9y agoTo summarise: they're not going to send which URLs you visit. However, they can set up specific experiments, and will send the results. Thus, they could set up an experiment "did users first visit example.com and later example.com/profile/igorpartola, and then have only you report that to be the case. That example is of course not that likely and thus troublesome. What worries me, though, is the described experiment "what search terms have led to users switching search engine", and then reporting the search engine. They acknowledge this problem, but the response is "at least that's better than gathering that data server-side". True, but it's still too much data, I think.