Stefan merged this earlier today and it should make the release now.

Another PR for fixing UI integration tests: https://github.com/OpenNMS/opennms/pull/3613

PR: https://github.com/OpenNMS/opennms/pull/3592

The existing classification engine does only include some optimizations for selecting rules with ports. In case that a rule base heavily relies on address matches rule selection is a linear search. In addition, the current rule engine does not support port ranges explictly. Port ranges are represented as corresponding single ports which is highly inefficient for large port ranges.

A new rule engine was developed that uses a decision tree to select rules. On each node a threshold for one of the aspects (src addr, dst addr, src port, dst port, or protocol) is used to split the rule set. For classification the decision tree is traversed and finally the classification rules found at the leaves are applied.