Huffman Coding

An important requirement for a Huffman encoder and decoder to work together is that the code word dictionary, or lookup table, must be available to both encoder and decoder. Of course, the two must also be identical. There are two obvious approaches to this problem

One approach would be to rely on globally agreed statistics, and use the same code word dictionary for many communication sessions, either by communicating the code word dictionary once or have the encoder and decoder derive it based on the same probability data. What would be the advantages and disadvantages of this approach?

The second approach is to analyse the message (or file, or image, or whatever) we want to send, obtain the probability statistics for our symbol set, generate the optimum code word set for this data set, and send the dictionary along with the message. What would be the advantages and disadvantages of this approach?