Unfortunately, too many users failed in their first attempt to use the 8255, hence, they typically dislike the device as being "bad" in some way. In reality, it provides ports and, in some cases, "features" not otherwise available, but all they remember is that they were unable to make it work. It's a simple reality.

Back when I was starting out, often using commercial-off-the-shelf (COTS) boards in many cases, the 8255 was a common feature on these boards. Unfortunately, because the 8255 couldn't drive a typical load, it had to be buffered, and the buffering that these boards provided, while helpful if it worked out properly, often didn't support all combinations of features that the 8255 was supposed to provide. Sometimes the buffers applied to the data but not the handshakes. Sometimes the buffers were bidirectional, in nybbles, which limited the wa in which the signals could be used.

With the limitation on handshaking via Port C that are placed on the device by virtue of its original design, and absence of internal register resources by means of which one could make those functions more general, I'm not certain that an 8255 would be worth putting out on the replacement market, even if you could speed 'em up by an order of magnitude and increase the Ioh/Iol to at least 6 mA, simply because it is limited. If you can't incorporate the address latch, because you have to retain the pinout, (perhaps you can make a 6824 version that fits) you won't have additional register space.

A glue-less part that just attaches as part of the expanded memory bus might have a place. Just another 8255 wouldn't be that appealing.

RE