X3T9.2/90-062







April 19, 1990

To:       X3T9.2 Committee
From:     John Lohmeyer, NCR Principal Member of X3T9.2
Subject:  Initiator Control of Reselection Order

Some SCSI applications require that the initiator have more direct control 
over the order in which reselections occur on the SCSI bus.  The present tools 
in SCSI-2 give some measure of control, but they are insufficient for the disk 
array applications that we envision.

While we have considered several other methods, and are open to alternative 
suggestions, we feel that the best way to solve this problem is to add two new 
messages.  One of these messages would create a means for initiators to reject 
unwanted reselections indefinitely, without aborting the I/O process.  The 
second message would permit the initiator to reselect the target when the 
initiator is ready to continue the I/O process.  For the sake of discussion 
only, I suggest the following names for the messages: DO NOT CALL ME (DNCM) 
and CONTINUE.

The DNCM message would be permitted in two places:
  1)  Following the IDENTIFY message (and any queue tag message) in the 
initial connection of an I/O process.  The target may continue the I/O process 
up to the point where it normally disconnects.  After that point, the target 
does not reselect the initiator.  Instead, the target waits for the initiator 
to reconnect via a selection, IDENTIFY, [QUEUE TAG], CONTINUE message 
sequence.
  2)  In response to a target reconnection, the initiator could send a DNCM 
message.  The target would respond with a DISCONNECT message and go to BUS 
FREE phase.  The target would not attempt to reconnect for the I/O process 
until an initiator reconnection occurs.

An initiator would reconnect to an I/O process via a selection, IDENTIFY, 
[QUEUE TAG], CONTINUE message sequence.  The function of the CONTINUE message 
is to avoid confusion with an Incorrect Initiator Reconnection.  The CONTINUE 
message would be added to the list of legal initiator reconnections in item 
(3) of the first paragraph of 6.5.2.

There are still some details that need fleshing out in this proposal and I 
will be glad to write a more complete proposal for the working group if this 
direction is accepted by the plenary.