>>540>>571The difference between analytics and surveillance starts with goals and ethics.
What are you trying to measure? The number of bits in the track list of
>>551 is clearly more than enough to uniquely identify all the users here, if not all humanity.
This new lainchan with analytics you now have is clearly unethical. Not only you collect all of
>>551, but you also submit most of that information to google via js and fonts. I'm fine on my Tor Browser with everything third-party off by default and most of those values you measure spoofed. But why would you subject less sophisticated users to surveillance on a board that purports to respect /cyb/ values, including, one would assume, privacy and anonymity, I can't possibly understand.
This is going to be my last informative post here until you reconsider, because now I have no other choice to continue being ethical to other lains than to be as benign as possible so that anything I say couldn't possibly damage someone else's police record. Which Google kindly collects for you right now and a guy that will hack into your server or you yourself fucking the server config up would provide to NSA later.
Somewhat more ethical setup would be:
>User IP addressMust not be recorded. Should be converted to a country code/"tor exit" with geoip and tor exit nodes list. Having an .onion would help too, btw.
> Date and time of the requestMust be rounded to 15-60 or more minutes depending on the amount of traffic to the board. No single user should be identifiable by post time.
> Title of the page being viewed (Page Title)Fine.
> URL of the page being viewed (Page URL)Fine on lainchan, but most of the time should be cleaned up of ids and other soykaf.
> URL of the page that was viewed prior to the current page (Referrer URL)Must be truncated to domain name.
> Screen resolution being used> Time in local user’s timezone> Links to an outside domain that were clicked (Outlink)Must not be recorded.
> Files that were clicked and downloaded (Download)Tracking links to outside domains is unacceptable.
Local links might be ok, depending on a threat model. Consider that raiding a given lains' computer with downloaded PDFs and pics can be used to link that lain to at least some posts.
Aggregated per URL popularity is fine.
> Pages generation time (Page speed)> Location of the user: country, region, city, approximate latitude and longitude (Geolocation)Country is ok, nothing else must be recorded.
> Main Language of the browser being used (Accept-Language header)What for? Must not be recorded.
> User Agent of the browser being used (User-Agent header)One bit (mobile/desktop) is fine. Everything else is unreliable for sophisticated users anyway.
OS type (Linux, Windows, OSX, other BSD) might be ok (but unreliable).
Nothing else must be recorded.
> No third-party resources or trackers.Obviously.
Truly ethical setup would be to just directly compute counters instead of recording any request data at all and throwing away any counters with less than 50 requests/day so that, for instance, users from Iran can't be identified just by correlating your per-country counter for Iran with some logs Iranian police already has. This is what tor daemon does for relay stats, btw.
> Countly/"tor exit" distribution.> URL (thread) popularity.> Load/hour/time of day.> Referer domain popularity.> Mobile/desktop popularity.Are both useful to you and not invasive for lains if you respect the 50/day rule above.
