A. Stationary targets
B. Moving targets
C. Patterns of motion (repeated visits averaged over time)
As earlier versions of geopriv-policy show, "class A" algorithms that only address the first can be pretty simple. As you try to address the other two, you end up with something like the most recent version of -policy or draft-thomson-geopriv-location-obscuring.
I would like to suggest that the base policy document doesn't need to address all of these problems. Advanced "class B/C" algorithms are clearly an active area for research and development, and the algorithm that the server applies is not a requirement for interoperability. The policy document format only describes the user's preferences; the server can reject policies or warn the user, but that's related to the policy *protocol* and the server's local policies.
A more concrete proposal:
1. Roll back the fuzzing algorithm to a simple "class A" algorithm
2. Document how this algorithm can fail in "class B/C" situations
3. Provide informative references to more advanced strategies, such as draft-thomson-geopriv-location-obscuring or a new draft with Jorge's algorithm.
Thoughts?
--Richard
_______________________________________________
Geopriv mailing list
Geopriv@ietf.org
https://www.ietf.org/mailman/listinfo/geopriv
