Does anything keep this from being a global unique ID?

[michaelkleber]

Your proposals says

1. Domains may elect to federate themselves together into a first party set operated by a single gatekeeper

and, once they've done so,

Browsers should then allow members of the set to pass a common user id set in a browser storage container between themselves.

Does that mean that if all the domains in the world decided to be in a single set, every person would get a unique ID that followed them across the entire web?

Your business rules also say

5. Once federated, the members of a set may share the unique id with one another & to a gatekeeper for cohort creation

(where I added the emphasis to the last few words). So maybe you don't intend this to be an ID that could be used for tracking, but somehow one that could only be used to create cohorts? That's what I was expecting to see after hearing your brief overview at the Web-Advertising BG. But I don't see what in this proposal keeps the ID from being a global tracking vector.

[jdwieland8282]

Hi @michaelkleber, the Gatekeeper would salt the userid before sending it to pub A,B...Z, such that when it is returned to the Gatekeeper only the Gatekeeper knows it's the same user. I see now that that isn't very clear in my diagram I'll fix that now, thanks for the feedback.

Does that mean that if all the domains in the world decided to be in a single set, every person would get a unique ID that followed them across the entire web?

We still need to define who gets to form a set, I don't think allowing all domains in the world to federate into a set is the right approach, nor do I think ownership of domains is right. There is a lot in-between those two extremes. Seems like an opportunity to compromise.

[jdwieland8282]

Does https://github.com/MagniteEngineering/Gatekeeper#gatekeeper-declaration-new-user help clarify?

[michaelkleber]

I understand the idea now, thanks.

(Please consider moving the step-by-step explanation into real text, instead of having it part of the image? Text inside images isn't accessible to screen readers, isn't something another person could propose edits to with a pull request, etc.)