Ah, but if you put a board which (like the PCI tulip and DEFPA FDDI)
*does* need bus-mastering, into a PCI slot which doesn't support
bus-mastering, you've already lost.  And this mostly (solely?)
affects early 486 PCI implementations, anyway.

Isnt the tradeoff really: Are we better off to work around buggy
BIOS/setup which dont enable the busmastering bit, or to leave it bit
alone, and risk not catching the `board incorrectly inserted in
non-busmaster slot' case?