The FLoC debate is pretty binary - you're either for it or against it. I think it's better to frame the debate as "how much tracking entropy should browsers pro

The FLoC debate is pretty binary - you're either for it or against it. I think it's better to frame the debate as "how much tracking entropy should browsers provide?" Tracking entropy is log(cohort you're in). So if a service can tell you're in a group of 1024 users, tracking entropy is log2(1024) = 10. The cohort you're in currently determined by 1) third-party cookies 2) fingerprinting techniques. Removing third-party cookies and introducing FLoC will probably reduce the entropy provided by the user. Recall that the FLoC proposal aims to put each user in a group of several thousand other users. That's about 12 bits of entropy. A third-party cookie would probably provide more, though I don't know the number off the top of my head. You only need log2(3 billion internet users) = 32 bits to identify every internet user hyper-precisely. So, moving to FLoC probably reduces the tracking entropy provided by the user. But it still leaves fingerprinting as a viable way to identify users. Even if both third-party cookies and FLoC were eliminated, there would still be fingerprinting. So, I think the Google approach is "provide a minimum tracking entropy via FLoC, and try to bound maximum entropy by limiting fingerprinting." Privacy advocates want a world where browsers try aggressively to limit tracking entropy, perhaps ideally eliminating it altogether. See the "privacy budget" mentioned here for a similar idea: https://blog.chromium.org/2019/08/potential-uses-for-privacy-sandbox.html https://blog.chromium.org/2019/08/potential-uses-for-privacy... Disclaimer: I work at Google.