When the functionality for "RESTRICTED SESSION" changed in 10g, a separate listener was required for accessing the restricted database instance from a remote host (with the "main" listener stopped). Having a separate gigabit network from the main fast ethernet LAN, where the SDU/TDU is set to a larger size to support database links (although jumbo ethernet frames do not work due to incompatibility between vendors packet sizes - the mostestest^^2 stupidest^^2 thing I've seen in years) we have a separate listener setup. (reminds me of the 1,000,000 vs 1,048,576 discussion not long ago). With being able to groom the listener log file via the SET LOG_FILE command in lsnrctl, I don't see a need for running 2 listeners just so that one can be cycled to groom the log. Where we have shared servers configured with dispatchers to handle VPNs with few known ports, we use a separate listener. In the case or where a client is going to use response time of generating 200 application logons simultaneously as a performance metric ... you might think about running as many listeners as CPUs (I'm not kidding). Paul