How are we thinking about PII?

open-feature / spec

OpenFeature specification

https://openfeature.dev

Apache License 2.0

661 stars 34 forks source link

How are we thinking about PII? #64

Closed justinabrahms closed 2 years ago

justinabrahms commented 2 years ago

We should be careful with Flag Evalutation Context, so that we don't shove tons of PII data there for cases where that's not needed.

toddbaert commented 2 years ago

A few wild ideas:

We could create hooks that would sanitize or warn about PII, and ship them. In the hook constructor, you might configure what fields or data patterns constitute sensitive, and then the hook before would take care of the rest.
We could also add switches to the evaluation options that would essentially remove all fields considered "sensitive".

beeme1mr commented 2 years ago

Perhaps we could do something like what LaunchDarkly does.

https://docs.launchdarkly.com/home/users/attributes#configuring-private-attribute-settings-in-your-sdk

Basically, LD allows you to optionally mark attributes as private. I believe they can still be used during flag evaluation but not they're not captured in analytics. @InTheCloudDan, can you confirm this behavior?

beeme1mr commented 2 years ago

The evaluation context spec now only contains targeting key as a recommended property. It's now up to developers to decide what additional properties they would like to include. @justinabrahms, do you think that sufficiently addresses your concern?

justinabrahms commented 2 years ago

Yep. I'm happy enough for now. :)