Hello Julian,

Sorry for the delay, I've been on holiday/vacation for a
week.  :-)

  JC> The table is correct - the mapping between "Labeled"
    > and "New device" is the one you want.

Thanks, that's good to know.  I still haven't found my TT
schematics, but I did unearth MegaSTE drawings, which may be
similar where the serial ports are concerned.

  JC> I used to have a 33.6K modem connectect to Modem2 and
    > run the connection between TT and modem at 38400.  I
    > can't remember the interrupt rate when it was
    > receiving at full speed, but I seem to remember it
    > using over 50% CPU.

38.4 KBpS is a useful enough speed: adequate for a 28.8 KBpS
with MNP4 enabled, or a 33.6 KBpS one with MNP4 turned off.
MNP5 (or perhaps V42bis) might be an option on a 14.4 KBpS
or slower modem.

The processor overhead might not be a problem with a simple
terminal emulator program, but it sounds as though anything
as hefty as a graphical Web browser is likely to suffer.

  JC> The problem is getting incoming characters out of the
    > 3-byte fifo on the chip and into the software buffer.
    > I did save some stats for the same setup on a Sun
    > 4/330 (similar ZS8530 serial chip) which show it
    > running at roughly 8000 interrupts/second (59%
    > interrupts on a 25Mhz CPU).

Hmm... will my 4/65 suffer the same problem?

  JC> If you run `sysstat vmstat`, you can see the interrupt
    > rate and CPU usage.

Thanks, that's a command worth noting.

  JC> I hunted down my TT manual last night.  The missing
    > lines are:
    >
    > Modem1                 6 (Data Set Ready)

I think that's missing on most of the Ataris I've looked at.
If I were connecting a serial terminal to a NetBSD TT, would
it (getty?) use DCD to detect the presence of a terminal?

  JC> Serial1                1 (Carrier Detect)
    >                        6 (Data Set Ready)
    >                        8 (Clear To Send)
    >                        9 (Ring Indicator)

That seems an odd combination of signals to omit, but I'm
sure it made sense to someone at Atari.

  JC> On Serial1, pin 4 (DTR) and pin 7 (RTS) are alwyas
    > asserted when the TT is powered on.

Doing that with DTR might make some kind of sense, but I
think it's a shame that RTS was hard-wired too.  It looks as
though the only option for Serial1 is software flow control
and blind assumption that both devices are ready.

Does the LAN port lack hardware flow control signals
(HSKo/HSKi or RTS/CTS) too then?

  JC> On Serial2, pin 9 is conencted to "Synchronous Clock"

Is that an output from the TT?  RI may not be a big loss, I
think most Hayes-style modems can indicate reception of a
ringing signal with a string.

Thanks for taking the time to reply, you've given me a much
better understanding of the TT's serial ports and their
capabilities.

- Andy Ball.