Closed miketaylr closed 2 years ago
Hmm, how would clients know how many people are nearby to produce a matching hint?
Good question - I don't really have anything very concrete. My high level idea is a user-agent would need to build a service that takes location (derived from network IP, GPS chip, etc.) and sticks it into a black box that knows about population density (from sources such as https://dataforgood.facebook.com/dfg/tools/high-resolution-population-density-maps). The service then gives a hash that fits the boundary conditions which can be sent to requesting sites.
Moving to geo IP entries: https://www.ietf.org/archive/id/draft-pauly-httpbis-geoip-hint-00.html
It would be useful if the geohash client hint operated over population density (or we could think of this as k-anonymity) rather than geographical distance from a given lat,lng coordinate pair. You could imagine specifying a default policy of some number, e.g., 500k people and provide a mechanism for sites to request more or less precision over density (lower bound could be UA policy defined).
Client hints don't have the concept of parameterized requests today, but they're defined as
sf-token
s, which allows for them. So possibly the CH could be extended to take ap
parameter that takes a value from n to 32, representing density of 2^p. For example,accept-ch: sec-ch-geohash; p=15
would be a request for the location corresponding to 32,768 people near me