I've been having troubles understanding this command and the frame-relay interface-dlci commands, but I think I've finally figured them out.

My question is -- when I was googling to try finding directions on setting up the frame-relay switch for labbing, the directions I ended up following were Chris Bryant's. One of his commands turns off inverse arp (necessitating the use of the frame-relay map command on the routers, if my understanding is correct), and his comments (on his youtube video at
YouTube - CCNA CCNP Frame Relay Switch Config 2/2 (it's at 3:23) say he does that command 'out of habit'. I'm curious if there's any reasoning behind his preferring to manually map the DLCIs? Odom's book said in a production environment you'd probably just let inverse arp take care of the mapping.
[EDIT]
ok; now that I'm labbing I find I don't understand the 2 quite as well as I thought.
When not using subinterfaces (meaning you have a full-mesh with common subnet) you can just map the DLCI's to the destination IP.
When using point-to-point subinterfaces you use the frame-relay interface-dlci, since there's only 1 router on the other end of the subinterface you don't need to map the remote IP address to the appropriate DLCI. I'm guessing the FR switch sees the frame coming in on DLCI 105 and throws it out the appropriate serial port based on it's settings (although this leads me to question the necessity of the frame-relay map command, if the FR switch can route correctly based on the DLCI alone).
When using multipoint subinterfaces, Odom's book uses the frame-relay interface-dlci command also, but I'm wondering why you couldn't use the frame-relay map command there, since they seem to be acting more like the first (non-subinterface) example - with multiple "routes" out that (sub)interface. Is it only for consistency? Physical interfaces use map while sub interfaces use interface-dlci?