It would be cool if we could implement some version of per-client metering. Under this, il-behaved clients would get locked out while good citizens could continue to use the endpoint. There are a number of challenges here:
- how to identify distinct clients
- how to ensure distinct responses per-client (ie: is there any caching that would get in the way of this)
- how to structure the per-client counter and make sure it doesn't get huge.
It would be cool if we could implement some version of per-client metering. Under this, il-behaved clients would get locked out while good citizens could continue to use the endpoint. There are a number of challenges here: