Sorry if there's a better list to discuss this on.

Platform's are i386 (2 - K6 266/Micronics M7STA-2B, FA310TX)

I'm baffled on this.  I run xntpd on one host and also run

  timed -M -F newhouse pimin

on "pimin" I run 

  timed

If I use timedc to check the clockdiff's between the machine's 
I notice that "pimin" is between 3000 & 8000 ms behind 
"newhouse".  Successive clockdiff's show that the gap is closing.
After ~1/2 and hour (I haven't measured this carefully it just 
seems like about that long) the gap begins to widen again.
I've let this sit for over 24 hours and the gap continues to widen.
"newhouse" stays really close to other remote hosts running xntpd.

If I restart either of the timed's I get this on pimin:

    Looking for nets to master
    Looking for loops
    readmsg: looking for ANY from ANY, ANYNET
    readmsg: wait 536.999994 at Thu Aug 20 00:44:55 1998
    
    Ignoreack: no ack: LOOP 1 1952   #10 205.219.89.41   newhouse
    readmsg: LOOP 1 1952   #10 205.219.89.41   newhouse
    slave ignored: LOOP 1 1952   #10 205.219.89.41   newhouse
    readmsg: looking for ANY from ANY, ANYNET
    readmsg: wait 515.999994 at Thu Aug 20 00:45:16 1998

Other net traffic between these two hosts and other hosts on this net
seem to be working ok (netscape, rlogin, rsh, qpopper, ssh, ...).

I'd really like to get it set up so only one host has to tie up my 
ISP connection with time syncing.

Am I doing something stupidly wrong with this?  

TIA,
Paul