If the other elements in the system provide a suitable environment (impedance relationships and gain structure), a transformer-based passive attenuator (using, for example, the Stevens & Billington TX-102) can provide fabulous results. I would go so far as to say it is the ultimate, in my experience. Caveat: in my system, I made certain that the gain structure and impedance relationships were suitable for a transformer volume control (TVC).
If this isn't practical in a particular system, the next step down is to use an active stage with a high quality resistive attenuator. A "passive preamp" using a resistive attenuator doesn't really provide the same sonic potential. The problem is caused by a relatively high output impedance at typical volume control settings. The typical result is decreased dynamics, HF extension, and increased coloration from the interconnect between the passive preamp and amp.
Is the potential sonic improvement elimination of a stage outweighed by the sonic penalty of a passive preamp's problems? In my experience, yes. Of course, this assumes that one uses a really good stage, and of course, a really good attenuator.
The best use of a "passive preamp" approach, as far as I am concerned, would result from locating the attenuator within the amp itself, minimizing the physical distance between the wiper and the input terminal of the amp's first amplifying device. But even there, I'd use a TVC if possible.
I have a financial (business) interest in a certain brand of passive stepped attenuator, but I use a TVC in my "ultimate" system because it is better still.