In article <199701201816.NAA03845@mimosa.com> you write:
> Is it easy to replace the UARTs in a 3/60 with a Z85230?  Does this
> cause SunOS problems?  Does SunOS exploit it?  Does NetBSD exploit it?

I've yet to try it, but it should be a simple plug in replacement. 
SunOS should automatically benefit from the larger receive FIFO, though
it won't take advantage of the deeper trigger depth setting.  Same for
NetBSD currently, plus NetBSD probably would benefit from the larger Tx
FIFO as well (since I think it writes as many bytes as it can to the
transmitter, so it should approach 1/4 the interrupt rate for
transmission)

NetBSD could be changed to set the chip to interrupt when the FIFO is
half full (4 bytes) instead of every byte in order to reduce the receive
interrupt rate to 1/4 as much as well, but the problem is that the chip
isn't smart enough to interrupt every so often by itself, so the driver
would need to install a timer interrupt to poll the chip every so often,
otherwise you could sit at a shell prompt forever having typed "ls\n",
while the chip waits for a fourth character to arrive. 

I think the 16550 is smarter in this respect, isn't it? I don't see
anything in the driver that looks like it handles this is doftware, so I
assume the chip will automatically interrupt every N character times or
whatever.