Charles> Markus Kilbinger <kilbi@rad.rwth-aachen.de> writes:

    >>  Hmm, I tried the patch, but it didn't change anything
    >> obviously. There are still lost chars on output decreasing
    >> dramatically the final transmission rate even at a serial speed
    >> down to 38400 baud.

    Charles> You're losing characters on *output*?  I've never heard
    Charles> of this before.  What's on the other side?  Does it have
    Charles> flow control enabled?

On the other side is a modem (max. 28800 baud) with enabled RTS/CTS
flow control. E. g. trying to download files from this machine on this
way (no matter using Zmodem or a PPP connection) leads to incorrect
transmitted data packages every 6-8 kB. Turning down the transmission
speed between this computer and the modem from 115200 to 38400 lowered
the error rate, but didn't solve the problem. We also proved that with
another modem and a null modem connected computer: Same behaviour. It
seems like the reaction on the hardware handshake fails sometimes as
the reason for this problem.

If it helps: At the beginning of this year there was a patch from
Sommerfeld (formerly available on ftp.netbsd.org .../arch/i386 and
some mirrors; I still have it ;-), but it's not (easily) applyable to
-current sources) which solved the serial problem completely for us.

Markus.