SSP meeting minutes #1 (1/94)

scheible at vnet.ibm.com scheible at vnet.ibm.com
Tue Jan 18 18:18:01 PST 1994


   These minutes also went out over the SSA reflector.  Please sign up for the
SSA reflector if you are interested in SSP.  Sorry for the duplication.


Accredited Standards Committee
X3, Information Processing Systems
                                               Doc: X3T10 / 94-017 r0
                                               Date: January 13th, 1994
                                               Project: N/A
                                               Ref Doc.:
                                               Reply to: John Scheible
To:             X3T10 Membership
From:           John Scheible

Subject :       Minutes of the January 1994 X3T10 SSP study group
                meeting (Meeting #1)


Agenda:

1.  Opening Remarks
2.  Approval of Agenda
3.  Attendance
4.  Document Distribution
5.  New Business
        5.1  Education
        5.2  Documentation format
        5.3  Alternative Pathing
        5.4  Auto Sense
        5.5  AEN
        5.6  Enlarged Split field
        5.7  Alternative Transport (ATM)
6.  Action Items
7.  Meeting schedule
8.  Adjournment




1.  Opening Remarks

   Since the SSP project proposal has not yet been formally
approved, the group met as a study group.  John Scheible, acting
whatever, called the meeting to order at 9:10 am on Thursday
January 13th, 1994.  Vitro is thanked for hosting the meeting.


2.  Approval of Agenda

   The agenda was created at approved.


3.  Attendance

   The attendees: (Please let me know if I misspelled your name).
        Ken Hallam              ENDL
        Stephen Finch           Silicon Systems
        Ed Haske                CMD Technology
        Erich Oetting           StorageTek
        Jeff Williams           Hewlett Packard
        Tak Asami               Western Digital
        Rene Macare             Future Domain Corp.
        Earl J. Perera          P.E. Logic Corp.
        Alan R. Olson           3M Co (observer)
        Edward A. Gardner       Digital Equipment Corporation
        Robert Liu              Fujitsu
        Bob Kembel              IBM
        John Moy                Tandem Computers
        Sam Karunanithi         Hitachi
        Ralph Weber             Digital Equipment Corporation
        Jeff Stai               Western Digital
        Paul Boulay             Hitachi
        Jerrieh L. Allen        Amdahl
        Clifford "Skip" Strang  BusLogic
        John Scheible           IBM


4.  Document distribution (no documents distributed, the SSA-SSP
    was in the last X3T9.2 mailing)


5.  New Business

        5.1  Education

      John Scheible spent about an hour doing an interactive
education session on SSA-PH and SSA-SSP.  Concerns were raised
that the Tag and LUN fields were too small, the vendor unique
fields were controversial (no consensus), and untagged queueing
may need to be supported.  The Split field was renamed OOT (Out
of Order Transfer) since split was misleading (Split=0) still
allowed segmented data transfer (split).

        5.2  Documentation format

   The format of the document was discussed.  In the last X3T9.2
mailing was the Proposed working draft SSA-SSP
(X3T9.2/93-0181r0).  John Scheible acknowledged that the
document was the SSA-SCSI (SCSI-2 mapping) document with a new
cover sheet.  It is recognized that the document will need to
undergo considerable work to make a SCSI-3 document.  A
discussion followed as to the proper format.  Jeff Tsai wanted
PageMaker(for Sun workstation people), ANSI wants Microsoft
Word.  John Scheible agreed to attempt to have the SSA-SSP
document updated and converted to Ami-Pro by next meeting.
There was a concern expressed that the import/export filters are
less than perfect.  John Scheible will investigate using Word at
a later date.  A discussion of tables and boxes not converting
to text files for Email was done.  Ralph Weber gave John
Scheible a C program to convert box characters to bars and equal
signs for text files, and John will investigate.  John will
remove the SCSI Status appendix, change the SSA-SCSI
references, remove the ACA write-up, and make the changes agreed
to later in the meeting, and re-release the document.  The
document will remain undergoing considerable change, and will
not be proposed to X3T10 as a Working Draft for some time.   The
terminology and services format will need to be changed for SAM
compliance.

        5.3  Alternative Pathing

   SSA-PH has the concept of alternative pathing which is not
handled well in SCSI.   We discussed the problem of out of
order-ness when using more than one path.  We discussed the
concept that an Abort message may only affect operations on the
path used by the Abort.  However, this still causes problems on
Clear_queue and Device_reset operations.  People will think
about this issue and discuss at the next meeting.

        5.4  Auto Sense

   Auto sense options were discussed.  The winning proposal was
to use the Data_ready and Data_reply messages with a bit in byte
1 indicating Auto Sense.  The auto sense Data_reply message
operates differently from the normal Data_reply message in that
a change in the byte_count results in less data being sent
rather than being followed by another Data_reply message to
transfer the remainder.  This allows the initiator to throw away
unwanted auto sense data.  The auto sense data transfer uses the
same tag value as the associated command, and will occur prior
to the SCSI_status message (which frees the Tag).  Multiple auto
sense transfers will be allowed.  Editor's note: as I am writing
this I realize that the status codes will need to be the same
for multiple sense data transfers.  We need to discuss this next
time.  Auto sense is allowed for more than just Check Condition
Status.  The ACA bit in the control byte of the CDB will still
control the resetting of the ACA.  A bit in the SCSI_command
message will enable auto sense on a command by command basis.
It was desired that the auto sense data transfer not terminate,
affect, or even stop the normal data transfer.    Stephen Finch
discussed the possibility of getting the status code early with
the auto sense data, but this was rejected due to the problems
of what happens if the early status code disagrees with the
SCSI_status status code.  John Scheible will document.

   Other options discussed included sending auto sense data
after SCSI_status and command complete (problem: when is tag
freed).  Other options included a special message with auto
sense data (problem: messages are only 32 bytes long, and would
require a set of them).  Other options included a pre-assigned
channel for auto sense data at SCSI_command time (channels can
be critical resources and could result in buffer overflow is
re-used).

        5.5  AEN

   AEN was also discussed.  Two options were discussed (Async
Alert, and Dara_ready/reply).  John Scheible did not like using
the Async_alert process used in SSA-PH because 1) all 255 bytes
of sense data could not be sent and 2) AEN is a SCSI concept,
not an SSA-PH concept.  It was realized that we should not be
using AEN as that is a particular SCSI-2 concept, but should
call it Asynchronous Events.  Editor's note: SAM calls it
Asynchronous Event Reporting.  The SCSI-2 method of AEN was
discussed, but it was decided that SSA may define a new method.
The concept of re-registering after each AE was discussed and
rejected.  The need to re-register in P1394 comes about due to
the possibility of buffer overflow.  In SSP, this does not apply
since the Date_ready / Data_reply process cannot overflow (the
initiator allocates the buffer space on the receipt of the
Data_ready before sending the Data_reply).  The AE data transfer
will operate like the Auto Sense transfer in that a Data_reply
that changes the byte count will truncate data transfer and is
not responsible for following up with another Data_reply for the
remainder.  John Scheible will write up for the next meeting.
The method of registering was not discussed.

        5.6  Enlarged Split field

   John Scheible proposed an enlarged split field to control out
of order transfer over the interface (1 bit) and over the media
(1 bit).  The problem is the device could optimize performance
of both reads and writes by performing out of order transfers on
the media, even if the initiator did not support out of order
transfers over the interface.  However, there is an age old
problem with Unix where the beginning and end of a record are
timestamped.  If the last part was written (end timestamp) and
the first part was written (begin timestamp) but the device
errored out before writing the middle, then an undetected error
would result.  Strong opinions were voiced that this is an old
problem and the out of order on the media bit belonged in the
Write CDB, and not in the protocol.  No one committed themselves
to the suggestion that a proposal for the Write command change
be submitted.

        5.7  Alternative Transport (ATM)

   Ed Gardner asked the group to consider the possibility of
defining SSP in such a way as to allow it to operate on both the
SSA and ATM physical layers.  ATM is a 48 byte packetized serial
interface.  Ed will teach an ATM tutorial next meeting.


6.  Action Items:

The following action items are recorded.
        1)  SSA-SSP in standard word processor format (March - Scheible)
        2)  Rename Split field to OOT (Out of Order Transfer) (March -
            Scheible)
        3)  POTENTIAL proposal on larger LUN fields (?)
        4)  POTENTIAL proposal on larger Tag fields (?)
        5)  Update SSA-SSP document (change SSA-SCSI to SSA-SSP, remove
            status appendix and ACA write-up)  (March - Scheible)
        6)  POTENTIAL proposal on SAM terminology and services model
            write-up (?)
        7)  Alternative pathing discussion next meeting
        8)  Write up Auto Sense, Asynchronous Event Reporting method in
            next document.  (March - Scheible)
        9)   ATM pitch next meeting  (March - Ed Gardner)


7.  Meeting schedule

   The next meeting will be at 9:00 am - 12:00 noon on Thursday
March 17th, 1994, and will hopefully be an approved project.
The May meeting time has not been finalized at this time.  The
meetings are subject to change, and everyone should watch the
X3T10 meeting schedule for the latest information.


8.  Adjournment

   The meeting was adjourned at 12:00 noon.



Sincerely,

John Scheible




More information about the T10 mailing list