On Sat, Jul 19, 2003 at 03:42:37PM -0700, 1 wrote: > I'm going to be writing a dissector for a custom protocol > that starts off by having a client connect to a server at a > well known port. The first response from the server > specifies two dynamically obtained ports. Both the client > and server close the initial connection and then the client > reconnects on the dynamically obtained ports.
The way this is generally handled in Ethereal is to set up a conversation, using the addresses and server port that will be used by the subsequent connection, and associate the appropriate dissector with it. (If necessary, the client address can be omitted if it's not known what address the client would use.) See, for example, the RTSP dissector.
