How or why Mozilla is involved (other than through a "We need money to survive yo" type arragement; and simply bundle the Pocket code for that reason alone, and whatever Pocket does is whatever Pocket does-- or if some other more nefarious deal is going on, I cannot say.
However, the Pocket code itself clearly does session-to-session data collection, as the material it curates changes based on what you click on. If it tracks anything outside of those "from the landing page clicks", I dont know. I have not tried to examine it under a microscope.
Also, not ENTIRELY apples to oranges for the quoted article.
The conditions presented by the article are:
Large database that contains your information inside it, but stripped of your unique ID.
This is functionally identical to "Large database of browsing habits, that lacks your unique ID."
Each instance of the pocket software does have a unique ID, but that is the ID of the pocket software installation, not of the profile being evaluated. If you have firefox installed on say, 5 devices, but you have the same browsing habits on all 5 devices, the same correlation techniques can be used to consolidate all 5 of those "unique" datasets into a single, unified dataset, with increasing levels of predictive accuracy, the more datapoints that are tracked and correlated.
That was the point.
That is entirely possible, because Facebook was ACTIVELY DOING THAT. That is exactly what a facebook "Shadow Profile" was.
https://www.cnet.com/news/shadow-profiles-facebook-has-information-you-didnt-hand-over/When you created a real facebook account, the backend would comb its shadow accounts to see if it found a close match, then ask some questions to verify-- then BOOM-- their shadow account (that they were indeed using for analytics purposes), suddenly has a real human ID with it.
But that's facebook. Their brazen disregard for actual user privacy is legendary. We are discussing Pocket and Mozilla. Pocket at least, must track "session to session, within the scope of a single instance ID, based tracking" in order for its curation feature to work. I dont really feel like setting up trail tests to see if they are aggregating and consolidating in this capacity, but it would be pretty easy to test for.