Should I handle very detailed subdissectors internally or externally with dissector table

Hi, forks, this is a best practice question. I am decoding a protocol set on top of UDP. The UDP payload is encapsulated in MotherProto, and the payload of MotherProto takes on several dozens of forms.

Each form represents a different message type, and is identified with UDP port number. For example,

udp port 64000 means A01 message, in which byte 1 is status, byte 2 is data_freshness, byte 3-4 is measurement_1
udp port 64001 means B02 message, in which byte 1 is status, byte 2 is data_freshness, byte 3 is command_1, byte 4-5 is command_2

...

There are two ways to handle this as I see it. One way is to set up a sub_dissector table "mothеr.port", and register each tiny little message type as a sub-dissector

Or, I can make call a unique subdissector in MotherProto, Childproto, and handle the message classification internally. One good thing about this is that I don't need to write so many handoff routines and I can put port-to-messageType information in a xml, which is easy to maintain and update (the correlation between port and message type may change between different releases).

Which is the best practice? Or is there a better/standard way? Mother and Child proto are both niche protocols unlikely to be encapsulated in any other way.

Thank you for your help!