That PR migrated code from message_data that handles an earlier (2020-12) edition of these data.
The data is (a) proprietary and (b) provided in an idiosyncratic, slightly complicated format. For the earlier PR, it was not worth the effort to produce a "fuzzed" (random values, but identical structure) version of the data for testing purposes.
Follow-up from #162.
To do: