Re: iscsi : default iscsi mode page settings - Consensus call

["Julian Satran" <Julian_Satran@il.ibm.com>]

Ralph,

That was the path and endlless-month mail exchange (with very little new
argument in each round) has brought us.
In summary:

   it was considered bad form to leave 2 layers (the layering argument) own
   (change) mode page variable
   we attempted to segregate the variables and argue that the burst-sizes
   are used by iSCSI but their values will be dictated by SCSI as they
   determine the multiplexing capabilities of the transport
   against this the argument was made that the mode page mechanism is hard
   to handle and a specific protocol has anyhow a mode page setup that is
   specific (that is not the current practice but is not explicitly
   forbidden by SPC)
   with all the above it makes little sense to support any mode page
   content (as SCSI can't control any value)

My question is - would returning the mode pages with a read-only content
make any-sense (considering that SCSI can't directly control the burst
values and that is the only reason SCSI may want them)?

The alternative would be (as I attempted) to leave both burst values under
SCSI control (where layering will require them to be) against which a
majority (of 2) brought the "very hard to handle" argument (which BTW has
some ground as text mode negotiations are easier than the set/enquiry ) and
reopen the consensus-call.

Regards,
Julo




                                                                                              
                    Ralph Weber                                                               
                    <ralphoweber@compu       To:     IPS Reflector <ips@ece.cmu.edu>          
                    serve.com>               cc:                                              
                    Sent by:                 Subject:     Re: iscsi : default iscsi mode page 
                    owner-ips@ece.cmu.        settings - Consensus call                       
                    edu                                                                       
                                                                                              
                                                                                              
                    01-10-01 17:56                                                            
                    Please respond to                                                         
                    ENDL_TX                                                                   
                                                                                              
                                                                                              



Santosh,

I was not aware that you were proposing that iSCSI
not support the Disconnect-Reconnect mode page at
all.  Now that would be a radical step.

If on the other hand you are proposing that iSCSI
have a Maximum Burst Size but refuse to provide
that information in the Disconnect-Reconnect mode
page whose page code iSCSI devices accept, then
I think you are at least violating the ISP charter
as regards interactions with T10 standards (specifically
SPC-2).

This is my rather strongly held opinion and unless
you find an argument that is substantially better
than anything presented to date I expect to maintain
this opinion to the bitter end.

Ralph Weber