There should be a trait for describing a "decoder" for data in a connector. A decoder comprises

• an initialization function, called once on pipeline startup
• a decoder(&[u8]) -> Result<Vec<u8>, Error> function

The configuration for a connector will specify the name of an encoder. The user has to implement this trait for the encoder in the udf.rs file. The decoder function will be hooked to be executed between transport and parsing, and will pass the result to the parser for further processing. The design could potentially allow for a stack of such decoders for each connector.