That's the goal. The pattern-of-life[1] analysis that GA enables is an incredible long-term risk.
Allowing one entity to log almost all of your internet activities into one aggregated database creates a massive amount of potential-social-power. Even if Google isn't currently converting that potential-power into an abusive form, the very existence of that data in one location creates a huge temptation for thieves and governments with national security letters.
> roll their own analytics
Server logs only give you a view of the traffic to one website. There is a huge difference between one business knowing you use their service and a 3rd party carefully logging your entire daily schedule.
[1] I highly recommend reading http://labs.rs/en/metadata/ which unfortunately didn't get much attention when it was recently posted to HN. I suspect most people are still thinking way too small when judging what is possible from "just metadata".
It adds plenty of protection for the user to block it.
I have no idea how people can look at the demographic information GA delivers and then conclude it isn't tracking users around the web...
I agree that the proliferation of tracker blocking could be an issue for smaller publishers and businesses in the near future. But as a privacy-conscious internet user, I'm excited about the vacuum this might leave for better self-hosted options.
Google analytics loads scripts from and sends results to Google. Those requests inevitably include IP address and referrer. That's more than enough.
What this does is it shows middle finger to Google, which is the tracking entity in the vast majority of cases.