At 17:44 Uhr +0200 31.7.2000, T&B wrote:

> 57600   modem cdtrcts  at&fe0     modem hung up on 3 attempts when ftp
>started

That's why you need "AT&D0". What happens here is that Macintosh "hardware
handshake" modem cables connect HSKo (the only handshake output line the
Mac has) with RTS (flow control) *and* DTR (line control). If you want flow
control, you have to explicitely ("AT&D0") tell the modem to ignore the
state of its DTR pin. Otherwise, the Mac lowers its handshake output ("Hold
on for a moment..."), but for the modem line control ("Hang up NOW!") takes
precedence over flow control.

I.e. you may be fine for some time without &D0, but during the first long
data burst your modem will kill the connection.

[...]

>But this best situation still only does 3.6 kB/s, not the 5.4
>provided by a Mac OS based router on similar hardware.

The throughput you will get is limited by tty speed. Try 57k6 with hardware
handshake, and you will see a performance like in MacOS.

>The "o/flows" are the sum of the silo number in these errors in the
>message log:
>Aug  1 00:09:02 macbsd /netbsd: zstty1: 8 silo overflows, 0 ibuf floods
>Should I be concerned about these?

Yes. For each of those "silo" (really: serial port fifo) overflows, a whole
packet has to be resent (500..1500 bytes). An occasional silo overflow
during high load is no problem, but if you see those regularly on the
unloaded machine, you'll be better off with a lower tty speed.

>Is there any way to achieve those higher serial speeds without errors?

Different hardware. Your IIcx has to process an interrupt for every
character it gets because the SCC's fifo is only three bytes deep. Other
ways to deal with the situation would be DMA (like in most PowerMacs) or an
I/O-controller like the CS Hustler cards (for which there is no driver in
MacBSD).

	hauke



--
"It's never straight up and down"     (DEVO)