Oops; I overstated. The end device should be PRESENT if the termination is to be correct, but the UDMA document I was thinking of (ATA-ATAPI 5 rev. 3) only requires the master at the end in section 5.2.13.2 saying "For ... all 80-conductor cable assemblies... device 0 shall be at the opposite end of the cable from the host" and this relates not to the termination condition but to the cable-select feature.
The termination of master and slave (device 0 and device 1) is not different in the electrical sense.