I fully support your comments. It is really up to Kafka to see whether such a simple approach would be sufficient. This probably also depends on the sophistication he can expect from his users (the more sophisticated they are the more likely this approach will not work). Whether or not testing against data is useful, would also depend on the context. If the data is from a limited domain, it could work. My suspicion though is, that he wants to do the checks to avoid unnecessary matching against data...