Norm,

     Re late errors:  on writes, a late error will always
cause a partial write because it is too late to stop the
pfsms from sending data.  I agree that treating all late
errors as partial transfers is reasonable.

     Of course you are right about the "long transfer" bit
indicating a transfer of >= 65,536 words.

     The experiment which is supposed to generate a timeout
now works.  This problem was a result of our connecting
the lsout signal from the cam chip directly to the node's
rs1in input, rather than using the appropriate multipurpose
pin.  I fixed it by adding a tristate buffer for lsout and
enabling it with the modsel signal from the cam chip.

     The lack of feedback from the cam.edf.v file is my
fault; I commented it out while I was debugging because the
displayed info was slowing me down.  It's back in now.

     Regarding the delay between Sbus grant and request,
we now have a three clock delay from the rising edge of
grant to the falling edge of the next request.  I will fix
this in the next update.

     On the extra burst subject, it's certainly true that
the extra bursts will have a very small effect on long
transfers.  Shall I remove the hardware?

                           Doug