In message <19980119145902.09316@antioche.lip6.fr>, Manuel Bouyer writes:
>I'm really happy with my aha2940UW. I don't know if they make
>differencial adapters based on the same chip ...

They do, it's the 2944UW.  Those are the only two listed in:
http://necxdirect.necx.com:8002/cgi-bin/auth/:guest/simple_q?cifilelnk=0000102
I am guessing some of the other, high-performance controller makers
might also make these cards, but I don't know of any.

I was also quite happy with my old AHA-1542B.

However, I have somewhere in the back of my mind a tick-mark against
them, but I can't remember what it was for.  Perhaps it was for being
overpriced (compare to the NCR-based cards), or for not moving to CAM
(see http://www.ultranet.com/~gfield/gary/scsigame.txt), or for going
to the software-configurable-only cards, or for being hard to get in
touch with (I think someone said it was "like trying reach someone at
IBM"), or because I heard they weren't releasing programmer specs, or
perhaps it was some second-hand information I heard about the 294[04]UW
in the Linux newsgroups.

Wait a minute, memory flash; is the 2940 the card which involves
downloading some code at boot time to the embedded processor (or
is it the sequencer engine), and they consider(ed) that protocol
proprietary, and the free Unix people had to reverse-engineer it?

Can anyone confirm this?
I wouldn't want to support programmer-hostile business practice.

Still, your datapoint is relevant and useful.  Thanks.