@ October 31, 1987 To: X3T9.2 Membership From: John Lohmeyer, X3T9.2 Chairman and Larry Lamers, General Editor Subject: Orlando Working Group Minutes A working group meeting was held October 26-28, 1987 at the Grosvenor Resort Hotel in Lake Buena Vista (Orlando), Florida. We would like to thank Steve Goldman of DPT for hosting the meeting and the editors working session on October 29-30. The following people attended the meeting: Name Status Organization ------------------------------ ------ ------------------------------ Mr. Robert N. Snively P Adaptec, Inc. Mr. James J. Semenak P AT&T Information Systems Mr. Paul Kellogg O AT&T Information Systems Mr. David Harms O AT&T Information Systems Mr. Gene Milligan P CDC/MPI Mr. Gerald Houlder A CDC/MPI Mr. Dan Davies P Cipher Data Products, Inc. Mr. Steve Goldman P DPT Mr. Tom Treadway A DPT Mr. Dave Kuszmaul S DPT Mr. I. Dal Allan P ENDL Mr. Robert Liu P Fujitsu America, Inc. Mr. Pat Walker P Hewlett Packard Co. Mr. Greg Floryance P IBM Corp. Mr. Gary R. Stephens A IBM Corp. Mr. Robert Johnston S LMSI-CPI Mr. Bruce Sketter S LMSI-CPI Mr. John Lohmeyer P NCR Corp. Mr. Dave Skinner ST NCR Corp. Mr. James McGrath P Quantum Corp. Mr. Dexter Anderson P Scientific Micro Systems Mr. Dean Konell P Siemens Information Sys. Mr. Lawrence J. Lamers A Sony Corp. of America Mr. D. W. Spence A Texas Instruments Mr. Peter Dougherty P UNISYS Mr. Jeff Stai P Western Digital The following new documents were distributed at the meeting: Document Doc Date Author Description of Document ------------- -------- --------------- ---------------------------------------- X3T9.2/87-131 10/22/87 P. Boulay MODE SELECT Implementations to Support Rev 2 COPY X3T9.2/87-137 10/23/87 R. Snively Programmable Control of SCSI operating Rev 2 mode X3T9.2/87-154 10/11/87 D. McIntyre Clarification of Synchronous Data Rev 2 Transfer Negotiation X3T9.2/87-156 10/26/87 J. McGrath Reordering of Commands in Command Rev 1 Queuing X3T9.2/87-157 10/25/87 J. McGrath Providing Initiators with Zoning (or Rev 1 Notched Drive) Information X3T9.2/87-173 10/1/87 D. Pope Write Same capability in SEND/RECEIVE Rev 1 DIAGNOSTIC vs WRITE X3T9.2/87-177 10/11/87 B. Spence Proposed Direct-Access MODE SELECT Page Rev 1 10h Modifications X3T9.2/87-179 10/23/87 B. Snively Explicit control of extended contingent Rev 2 allegiance X3T9.2/87-180 10/22/87 G. Floryance Format Progress Reporting Rev 1 X3T9.2/87-184 10/23/87 B. Spence Reconcilation of Sequential-Access positional terms - Step 2 X3T9.2/87-186 10/24/87 B. Spence Proposed SEND DIAGNOSTIC Command Pages X3T9.2/87-187 10/26/87 G. Stephens MEDIUM CHANGER DEVICE CLASS MODEL X3T9.2/87-188 10/22/87 B. Snively Explicit selection of target functions X3T9.2/87-189 10/10/87 G. Floryance General Comments on SCSI-2 X3T9.2/87-190 10/28/87 G. Stephens Deferred Error Information Bits The meeting agenda included the following items: 1. Denoting optional features -- Editors to pick an economial style. 2. Archival vs. update operation of WORM media. (Boulay) 3. READ LOG command. [87-43R2,-113] 4. CD-ROM command set proposal. (Lamers) [87-106] 5. Clarify initiator ability to select a target to ABORT a pending command after the target disconnects. [87-168] 6. Buffer full/empty ratios. (Boulay) [87-172] 7. IGNORE WIDE RESIDUE Message. (Goldman) [87-174] 8. Fast SCSI. (Schuessler) [87-138,-148,-175] 9. Compatibility of non-queued targets with queued initiators. (Snively) 10. Programmable SCSI operating mode. (Snively) [87-137R1] 11. Serial number location issue. [87-127] 12. COPY problems with MODE SELECT. (Boulay) [87-131R1] 13. Format progress reporting. (Floryance) [87-180] 14. Nonchangable parameter handling. (Houlder) [87-163] 15. "Write same" as a SEND DIAGNOSTIC page. (Pope) [87-173] 16. Extended contingent allegiance. (Snively) [87-179] 17. Target implemented options. (Nitza) [87-130] 18. Reordering queued commands. (McGrath) [87-156] 19. MODE SELECT command modifications. (McGrath) [87-122R2] 20. Returning zone information. (McGrath) [87-157] 21. RECOVERED BUFFERED DATA issue. (Appleyard) 22. Flexible disk pin 1 issue. (Schlitzkus) [87-115] 23. Target vs. logical unit addressing. (Snively) 24. Residues on deferred error handling. (Stephens) [87-190] 25. Optional Wide Data Transfer Option. (Masiewicz) [87-181] 26. Data Pointer final value. 27. General comments on SCSI-2 (Spence, Floryance) [87-189] 28. Multi-byte message handling. 29. Explicit negotiation for 8-bit wide bus (m=0). 30. Initiator termination of commands. 31. REQUEST SENSE polling. 32. Restore pointers - mandatory or not. 33. Sense/Status in SEND/RECIEVE DIAGNOSTIC 34. Sequential access buffered mode two 35. Deferred errors 87-19. 36. Handling of reserved bits in Identify message. 37. Tape glossary 87-184. 38. Direct-access MODE SELECT page 10h [87-177]. 39. SEND/RECEIVE DIAGNOSTIC pages. 40. Clarification of SDTR [87-154 R2]. Due to limited time, not all agenda items were discussed. The following is the result of the discussions: 1. Denoting optional features -- Editors to pick an economical style. Assigned to the editors. 2. Archival vs. update operation of WORM media. (Boulay) Deferred, Paul was not present. 3. READ LOG command. [87-43R2,-113] There was lengthy discussion on this command. Most of it centered around the addition of pages from 87-113 and new pages to be added. There was a strong need expressed for vendor unique pages; if the PF bit is zero then all the page codes are vendor unique. The CDB will be changed to a ten-byte format and moved to Group Two. 4. CD-ROM command set proposal. (Lamers) [87-106] Larry Lamers conducted a training session on CD-ROM format/structure Tuesday evening. Since Paul Boulay was not present and Larry Lamers had not had time to revise his proposal this item was deferred. 5. Clarify initiator ability to select a target to ABORT a pending command after the target disconnects. [87-168] Wording should be added under the ABORT message that it is permitted for an initiator to select a target that has disconnected, in order to send IDENTIFY, ABORT. The initiator must keep ATN asserted continuously for both messages. The discussion digressed into the topic of an orderly abort procedure that does return residue information about the aborted command. No one wanted to change the current ABORT message, but several people thought a mechanism should be defined that would allow a host to terminate an I/O and get information from the target on how far the I/O progressed prior to termination. Dan Davies, Gary Stephens, and Steve Goldman accepted an action item to define the requirements for such an orderly termination procedure. 6. Buffer full/empty ratios. (Boulay) [87-172] Paul Boulay's proposal was accepted. 7. IGNORE WIDE RESIDUE Message. (Goldman) [87-174] There was considerable discussion on how often this message should be permitted: 1) Once per DATA phase, 2) Once per selection, or 3) Once per command. Also considered was whether the initiator can also send this message and, if so, how often? The results of this discussion was that the message should be permitted once per DATA phase (IN or OUT) and the message is only sent from the target to the initiator (MESSAGE IN). If this message occurs after a DATA OUT phase, the initiator is expected to adjust the data pointer appropriately and to resend the ignored data if the data transfer is continued. The discussion digressed at one point to the idea of an early termination (by the initiator) of a write operation. Currently, SCSI does not include this concept. The proponents of this concept may generate a proposal. 87-174 was accepted except that the message code should be 24h, not 10h. 8. Fast SCSI. (Schuessler) [87-138,-148,-175] Since Jim Schuessler was not present, this item was not included into SCSI-2 R3. Jim still has an action item to generate a detailed proposal with suggested wording. This item was placed on the agenda for next plenary meeting. 9. Compatibility of non-queued targets with queued initiators. (Snively) This item stems from the question of what is meant by "maximum queue depths" of zero and one? After considerable discussion, it was decided to eliminate this field in the queuing page (bytes 4 & 5). The only agreement reached was that the field isn't particularly valuable. 10. Programmable SCSI operating mode. (Snively) [87-137 R2] It was agreed to incorporate this in SCSI-2 R3 so that it would get wider visibility, but a warning will be included to indicate that more work is needed. Greg Floryance wanted a strong warning. Bob Snively felt that the ROM/RAM conversion issue still needed some serious discussions. 11. Serial number location issue. [87-127] The Serial Number will now occur only in one place, VPD page 83h. Page 83h needs to be developed. 12. COPY problems with MODE SELECT. (Boulay) [87-131R1] There was a discussion on using 3rd-Party RESERVE as a mechanism for transferring the MODE SELECT parameters (that is, while the 3rd party reservation is in effect, the target would use the reserving initiators MODE SELECT parameters on behalf of the device to which it is reserved. There was general agreement that this might work, but most people wanted more time to think about it. 13. Format progress reporting. (Floryance) [87-180] This topic raised questions on how the REQUEST SENSE command works in the event that a CHECK CONDITION has not occurred. This was assigned to item 31 to avoid sidetracking Greg's proposal. However the proposal will need to reflect the results of that discussion. The proposal was accepted for incorporation into SCSI-2 R3, with modifications to use the sense data codes as proposed by Bill Spence for TEST UNIT READY, and a two-byte Format Progress Indicator field. There was a discussion of what to do if the whole defect list cannot be accepted by the host at the start of formatting. The conclusion was that the format progress reporting cannot start until the target has accepted the complete defect list. Gary Stephens raised a general question regarding how a target is to deal with parameter data and reporting an error detected in the parameter list. He felt that a target should not begin a command until it has verified that the parameters are all correct. Jim McGrath pointed out that the current SCSI standard only requires targets to verify the CDB parameters prior to altering media. 14. Nonchangable parameter handling. (Houlder) [87-163] There was a long debate on the merits of each method versus actual field use. Option 1 was selected as the best implementation. This is to be incorporated into SCSI-2 R3. 15. "Write same" as a SEND DIAGNOSTIC page. (Pope) [87-173] Dexter Anderson accepted an action item to draft yet another proposal for handling the write same function offering it both ways. This will go back to Dan Moczarny's original proposal and use a separate operation code, and incorporate some of the features Jim McGrath desires. 16. Extended contingent allegiance. (Snively) [87-179 R2] Bob indicated that 87-179 R2 needs another revision to correct some errors. There was a long discussion of contingent allegiance and how/why it exists in SCSI-1. Jim McGrath expressed a desire for having the an explicit set of rules for when the target would send an INITIATE RECOVERY message to limit the number of versions of firmware he would have to generate to satisfy customers. The sense of the working group was that this was not possible considering the total number of potential error conditions. Bob generated 87-179 R3 at the working group and it was accepted for incorporation in SCSI-2 R3. 17. Target implemented options. (Nitza) [87-130] The desire to have a method of selecting the options is not addressed in Paul's proposal. In the absence of a proposal, 87-130 page 1 and 2 were accepted for incorporation into SCSI-2 R3 in INQUIRY standard format which reports the devices capabilities, not necessarily its configuration. Bob Snively indicated some options are already programmable in the ECA proposal. A definition is needed for reporting UNIT ATTENTION condition when changes in INQUIRY data occur (such as loading microcode and CHANGE DEFINITION commands). 18. Reordering queued commands. (McGrath) [87-156 R1] Destined for an appendix. This will be thrown into the SCSI-2 R3. 19. MODE SELECT command modifications. (McGrath) [87-122R2] Jim did not worked on this because of general problems with reporting sense data information. 20. Returning zone information. (McGrath) [87-157 R1] 87-157 R2 will be forthcoming from Jim. It has been accepted and is scheduled for incorporation into SCSI-2 R3. 21. RECOVERED BUFFERED DATA issue. (Appleyard) Responsibility for this item was reassigned to Gary Stephens. Jim McGrath will develop a proposal for disk devices. 22. Flexible disk pin 1 issue. (Schlitzkus) [87-115] Deferred because of a potential conflict with the flexible disk standard X3.80-198X. Dexter Anderson was given an action item to check this out. 23. Target vs. logical unit addressing. (Snively) [87-188 R0] Jim Semenak, Greg Floryance, and Gary Stephens are in love with this proposal and desire immediate incorporation into SCSI-2 R3. Jeff Stai is having heartaches, torn between his old love, SCSI-1 and the new beauty on the block ("one's got the money, the other's got the fame"). If any process exists, process zero must exist. The commands to which process zero responds are limited to certain commands. Dal started having nightmares (daymares??) when he began to realize the hole opened by allowing vendor unique commands through and trying to deal with eight different processes some, all, or none of which may exist. Dal strongly favored bondage, and limiting the implementation. Gary Stephens proposed that only process zero be allowed and the commands limited to those listed in the proposal with the removal of the vendor unique commands. This was favored (two members were not in favor). The phrase "in the target, which shall be zero" was added to the definition of bits 2-0. Jeff Stai accepted an action item to look into adding the COPY command to the list. Jeff proposed that the list be slimmed down to just the INQUIRY and REQUEST SENSE commands. This proposal was unanimously favored and everyone kissed and made up. The editor's marriages were saved. As amended, this document will be incorporated into SCSI-2 R3. 24. Residues on deferred error handling. (Stephens) [87-190] This proposal, which only addresses sequential-access devices, was accepted to be added to the section on REQUEST SENSE information bytes. Gary Stephens typed it up the proposal and gave the file to the editors. This will be incorporated into SCSI-2 R3. 25. Optional Wide Data Transfer Option. (Masiewicz) [87-181] The working group tended to polarize on this issue. It may have been that John Masiewicz needed to be present to explain his proposal. Interpreting the wording of the document lead to some rather passioned discussion. Dal was acting as the public defender and did a good job of arguing the case for 87-181. This was deferred to the plenary meeting since Dal couldn't filibuster the working group into acceptance. 26. Data Pointer final value. Is the final data pointer value in the host adapter one byte beyond the last byte transferred? An implementors note needs to be added to the document indicating that this is not guaranteed. 27. General comments on SCSI-2 (Spence, Floryance) [87-189] Only the following items from 87-189 were discussed: 1.1 NOP message allowed as first message. 4.1 ok 5.1 Treat as 1st party reserve. Requires 3rd Party RELEASE. Maybe an implementors note is needed. 6.2 Introduced QPE bit. The other items should be added to the next working group agenda. 28. Multi-byte message handling. This item was not discussed due to time restrictions. 29. Explicit negotiation for 8-bit wide bus (m=0). Explicit negotiation for an 8-bit wide data path is allowed (i.e., m = 0). The revision 2 draft document implied that the only way to negotiate for an 8-bit wide data path is to use MESSAGE REJECT in response to attempts to negotiate for a 16-bit or 32-bit path. 30. Initiator termination of commands. See agenda item 7. No proposal was generated on this during the meeting. 31. REQUEST SENSE polling. This item has to do with what is returned in the event that REQUEST SENSE is received without a prior CHECK CONDITION. Some implementations always return NO SENSE in this case. Other implementations return the current status of the logical unit. SCSI-1 permits either of these implementations. For polling situations (such as FORMAT Progress Reporting -- see agenda item 13), we may need to nail this down. The only safe way right now is to alternate between TEST UNIT READY and REQUEST SENSE. We also have a document from Dennis Appleyard (87-152) that addresses this issue -- it was not discussed during the meeting. 32. RESTORE POINTERS message - Should it remain mandatory? The RESTORE POINTERS message is used primarily for error recovery. The current draft SCSI-2 standard says that the initiator must implement this message. Several people expressed surprise/concern about this requirement. This item was added to the agenda of the next plenary meeting. 33. Sense data and status in SEND/RECEIVE DIAGNOSTIC There are some very different philosophies on diagnostics. Encapsulated commands appear to cause the greatest divergence. Dal requested proposals. 34. Sequential Access Buffered Mode Two [87-98 R2]. This proposal was accepted for incorporation in SCSI-2 R3. Dave Skinner accepted an action item to generate a Buffered Mode Three proposal. Jim Semenak accepted an action item to generate a proposal on the use of BUSY status, since he has objections to the wording in the 87-98 R2 proposal. 35. Deferred errors [87-19]. The working group accepted this for incorporation into SCSI-2 R3 with a note that ripple effects may occur in other parts of the document. 36. Handling of previously-reserved bits in the IDENTIFY Message. Reference item 23. Several ideas were presented: MESSAGE REJECT message, CHECK CONDITION status, or bus free. 87-188 was modified to specify how to deal with this by referring the Bit 4 definition to section 5.2.1, ATN definition. 37. Tape glossary 87-184. This was accepted and will be incorporated. 38. Direct-access MODE SELECT page 10h [87-177 R1]. This item was not discussed due to time restrictions. 39. SEND/RECEIVE DIAGNOSTIC pages. This item was not discussed due to time restrictions. 40. Clarification of SDTR [87-154 R2]. Jim McGrath disagreed with the implication that this is what synchronous negotiation was meant to be in SCSI-1. There was still concern over the clarity of the wording. The document was accepted for incorporation in SCSI-2 R3. Bill Spence will attempt some editorial clarification and offer it to the next plenary meeting. [Bill's clarified wording was included in the SCSI-2 R3 document.] cc: Attendees who are not P, A, or O members @