@tsteur opened this Pull Request on April 29th 2020 Member

The purpose of the fingerprint is to put all tracking requests within the same visit into the same visit in the DB. A fingerprint is basically valid for up to 30 minutes after the visit ends (unless different visit length is configured).

By adding a random hash to the fingerprint we make sure when a user disables cookies, a visitor cannot be identified across multiple days and such no cookie consent should be needed.

This change should not really change behaviour for anyone AFAIK. In Matomo 4 we could remove the check for create_new_visit_after_midnight potentially. We cannot limit the fingerprint in that case as when someone has create_new_visit_after_midnight=0 and has cookies disabled, then a new visit would be still created after midnight since the configId would change.

The limited fingerprint will also not apply when enabling [Tracker]enable_fingerprinting_across_websites = 1 since different sites can have different timezones and therefore we can't add any date to the fingerprint.

refs https://github.com/matomo-org/matomo/issues/13655

@tsteur commented on May 5th 2020 Member

Should be ready for a review. If I see things right, nothing changes for users in the end but hard to say.

@sgiehl commented on May 5th 2020 Member

Haven't looked deeper into the code yet. But wouldn't changing the config id will have the effect, that every visitor will be a new visitor after the update? 🤔

@tsteur commented on May 5th 2020 Member

@sgiehl only when cookies are disabled. Otherwise not. The fingerprint as @mattab mentioned was always only used to look 30 minutes back. So it would only affect it when cookies are disabled, and a visitor was active in the last 30 minutes. Then one new visit would be created which should be fine.

@tsteur commented on May 18th 2020 Member

@sgiehl did you maybe have another look?

@sgiehl commented on May 21st 2020 Member

@tsteur I've reverted the test file changes and the tests seem to pass again. From my point a view the changes look fine now. Maybe you can also have a look at the changes from @diosmosis. Otherwise feel free to merge

This Pull Request was closed on May 21st 2020
Powered by GitHub Issue Mirror