Mozilla may require users to opt-out of sending telemetry from its Firefox browser because so few have opted in that it's hard for developers to get a good sample of what causes problems.
The idea of opt-out telemetry has sparked a pretty lively mailing-list debate (at the time of writing, 42 posts in just a couple of days, from 31 authors, on what's a moderately-obscure topic) about how to improve that data collection.
The rough consensus so far is that if it approaches the question right, Firefox could flip to opt-out – just so long as it doesn't become a stalker.
The solution for that, as the thread discusses, is to follow Google's lead and implement what's known as differential privacy has used by Google's project RAPPOR).
After initial complaint on the email list, Brit developer Gervase Markham (lead developer on Bugzilla) chimed in to emphasise that Mozilla is not planning to gather data “with no privacy protection at all”.
Most controversial is the idea that the data collection let Mozilla compile a browser-side “top 100” of sites users visit. As developer David Teller explains, there is a “considerable bias” between users who okay telemetry collection and those who don't, which makes it hard for developers to optimise the browser for “websites that matter to our users”.
The problem with just working from Alexa data is that site owners like Facebook try to present a contextually customised page. To measure how a browser is performing on those sites, one would want to measure performance for actual content that real users are seeing”, Aaron Klotz wrote.
It's possible to do that without a specific phone-home: for example, the telemetry could start with a list of likely sites, so instead of saying “user visited google.com”, what's collected is “user visited one or more of these sites in the last week”.
And apart from privacy, that thought is problematic to some, since it would risk skewing development to get the best performance and presentation for a relatively small handful of huge sites.
Mozilla is going to start with a user study before it takes any steps down the opt-out path. ®