This should make it easier to troubleshoot issues caused by accidentally using a non-conformant JSON format. For example, it turns out that Dart's default JSON format is not the format defined by the proto3 mapping spec but some bespoke format that uses tag numbers instead of names.

Previously, a server or client would accept such messages and just ignore unrecognized fields (which could be all of them). This could result in hard-to-troubleshoot errors, since the server would behave like it got a partial or empty request message.

This will cause the codec to complain about the unrecognized field name, which should produce more intuitive errors.

We do similar for the Protobuf binary format, too, disallowing unrecognized fields in the data.