Since the full-duplex mode has a simpler design than the half-duplex mode, and the former’s efficiency is higher than the latter’s, why do we still bother implementing half-duplex mode in the Ethernet...

Since the full-duplex mode has a simpler design than the half-duplex mode, and the former’s efficiency is higher than the latter’s, why do we still bother implementing half-duplex mode in the Ethernet MAC?

If the Ethernet MAC operates in the full-duplex mode (very common at present), which components in the design should be disabled?