> > Well my results made me wonder if this wasn't some kind of configuration
> > problem. My first thoughts were about the hw-handshake. I use hw-handshake
> > consequently on both my modem and TT.
> 
> I think the problems are caused by high DTE<->DCE baud rates. I can't
> check at the moment, but I think that > 19200 is loosing so many
> interrupts, that troughput gets close to zero. Especially the SCSI
> driver interrupt code seems to be blocking the machine in high IPL
> for relatively long time.
I thought that too first. I already decreased the time spend at an ipl>0
in the scsi-driver a while ago. A few weeks ago, I mailed Daniel a small
patch that made the scsi-driver (nearly) always execute at ipl0, that
didn't help him, so I suspected something else. Maybe I should send you
and or Michael the same patch. I don't remember anymore if Michael was
running on a TT or Falcon, as the Falcon still polls for EDMA, the problem
is probably worse there. Thomas and I already have done _something_ about
it, but Daniel reported problems with it, so I didn't consider it ready
for 1.1 . This also needs some testing.

Leo.