Should tcp_dissect_pdus() add a data parameter to allow "private data" that was passed to a dissector running over TCP to be used by the dissector itself?
I debated this when removing the pinfo->private data used to pass struct tcpinfo data in r53036. The only current dissector that uses tcp_dissect_pdus() + the "private data" from TCP is the NDMP dissector, so I didn't think it was worth it and just used p_add_proto_data(). I didn't add p_add_proto_data() in enough places and caused bugs 9393-9394 which is what opened this discussion. Now I'm leaning more towards having the data parameter.
Curious as to other opinions.
Michael