Closed andrewbwm closed 3 years ago
Do we always want to propagate write/delete/deny to all data assets when all replicas have been compromised?
We do not support any sort of majority voting, if any functional replica is available it is assumed to restore the data to the desired state.
We may wish to introduce steps to represent an attacker poisoning the replicas.
The replication behaviour was updated in #69 to a less pessimistic interpretation that better describes common scenarios.
Introduce information replication via an association with Data assets that serve as backup replicas.
If one replica is read all replicas are read.
In order to do a write, deny, or delete on a replica all the replicas must have been disrupted(written, denied, deleted, not necessarily the same disruption on all of them, but all must have been affected by one of the three). Once replication has been broken via the previously mentioned, write, deny, or delete, those steps propagate to all the replicas.