Under NetBSD 1.3.2 / i386:

Jan 31 21:22:05 ns pppd[18419]: Serial connection established.
Jan 31 21:22:06 ns pppd[18419]: Using interface ppp0
Jan 31 21:22:06 ns pppd[18419]: Connect: ppp0 <--> /dev/tty01
Jan 31 21:22:40 ns pppd[18419]: local  IP address 212.246.39.2
Jan 31 21:22:40 ns pppd[18419]: remote IP address 212.246.39.129
Jan 31 21:24:40 ns pppd[18419]: LCP terminated by peer (Peer not responding)
Jan 31 21:24:43 ns pppd[18419]: Connection terminated.
Jan 31 21:24:44 ns pppd[18419]: Exit.

I know this can be caused by a dropped PPP link - but why does it happen
while ftp is transferring data happily all the time?  If I start ftp,
this will start happening again at 4-minute intervals (dependent on the
LCP configuration, I guess).  If I stop ftp, the link stays up for days.

Also, when the ftp is running, even ping packets fail to pass the PPP link.
So do name server queries.  BUT, I can still use the remote machine
interactively over ssh and the very same PPP link.

If I kill the ftp, everything returns back to normal.  Any ping packets
sent while ftp was running will be replied to as soon as the ftp packets
stop flowing.

pppd has been killed and restarted at both ends, no help.

I would think ping and LPC packets would always pass the link, no matter
how many ftp packets were queued up?  This is slightly annoying because
I wanted to ftp a 60+ MB file, but it makes my PPP link die every 4
minutes and prevents named queries etc...

The client end of the PPP link is running NetBSD 1.3F, the server (and
router) NetBSD 1.3.2.  I never had this kind of problems under NetBSD 1.2.

Any suggestions?

  -jm