To: X3T9.2 Membership From: John B. Lohmeyer, X3T9.2 Chairman Subject: May 1988 Working Group Meeting Minutes Gerry Houlder of Control Data Corp. hosted a working group meeting May 9-11, 1988 in Minneapolis, MN at the Airport Hilton. I would like to thank Gerry and Control Data for hosting a very productive meeting (with Modem-friendly phones). The final agenda was as follows: Agenda Items for SCSI-2 1. Additional Sense Code resolution 2. [] bracket removal 3. Review of S2R4 4. Rev 4 Issues - Sections 7 and 8 [88-38] (J. McGrath) 5. WORM Device Specific MODE SELECT Page [88-44] (P. Boulay) 6. Pin assignments for high density connector B 7. ACCESS LOG proposals (Spence) [88-006, 88-053] 8. Long vs. Short BUSY [87-192R3] (Semenak) 9. RECOVER BUFFERED DATA command for sequential access devices [88-022] (Appleyard) 10. Residues on WRITE FILEMARKS commands (Davies) 11. MODE SELECT rewording on saving pages and nonchangeable parameters [88- 027] (Houlder) 12. Proposed definition of caching [88-030] (McGrath) 13. Wide Data Transfer Request (Goldman) 14. Embedded Disk Cache parameters [88-049] (McGrath) 15. Sequential Access Initialization Procedure (88-42) 16. MODE SELECT Clarification (88-43) 17. "User-friendly" MODE SELECT commands 18. Helical Scan Density Code [88-52] (Hess) 19. Peter Johansson's Suggestion re SDTR Messages Agenda Items for SCSI-3 50. SEND DIAGNOSTIC Pages [87-186] (Spence) 51. LOAD SKIP MASK command proposal [87-203] [87-217] (Floryance, McIntyre) 52. More that 8 devices on wide SCSI [87-206] (Harms) 53. Search Command modifications [88-002] (Stai) 54. Expanded RelAdr Bit [88-007] (Boulay) The following people attended this meeting: Name Status Organization ------------------------------ ------ ------------------------------ Mr. Robert N. Snively P Adaptec, Inc. Mr. James J. Semenak P AT&T Mr. David Harms O AT&T Mr. Dan Davies P Cipher Data Products, Inc. Mr. Gene Milligan P Control Data Corp. Mr. Gerald Houlder A Control Data Corp. Mr. Paul Hanmann A Emulex Corp. Mr. I. Dal Allan P ENDL Mr. Robert Liu P Fujitsu America, Inc. Mr. Bob Pentecost O Hewlett Packard Co. Mr. Gerald Marazas A IBM Corp. Mr. David A. Buesing O IBM Corp. Mr. Skip Jones A Maxtor Corp. Mr. Fintan Buckley V Micropolis Corp. Mr. Kim Ryal A Miniscribe Corp. Mr. John Lohmeyer P NCR Corp. Mr. Dave Skinner S NCR Corp. Mr. James McGrath P Quantum Corp. Mr. Dexter Anderson P Scientific Micro Systems Mr. Tom Gardner O Technology Forums, Ltd. Mr. D. W. Spence P Texas Instruments Mr. Jeff Stai P Western Digital P - Principal A - Alternate O - Observer S - Special Interest V - Visitor The following new documents were distributed at the meeting: Document Doc Date Author Description of Document ------------- -------- --------------- --------------------------------------- X3T9.2/87-192 5/8/88 J. Semenak Proposal on returning additional BUSY Rev 3 information X3T9.2/88-49 5/9/88 J. McGrath Proposed Read Caching Page Rev 1 X3T9.2/88-53 5/8/88 B. Spence READ LOG/ACCESS LOG/LOG SENSE/SELECT Pages X3T9.2/88-54 5/9/88 J. McGrath August 29-31 Working Group meeting Announcement X3T9.2/88-55 5/9/88 J. McGrath Advisory vs. Mandatory Bit in the Proposed Read Caching Page X3T9.2/88-56 5/9/88 D. McIntyre July 11-13 Working Group Meeting Announcement X3T9.2/88-57 5/10/88 J. McGrath Handling Unsupported Pages in MODE SELECT/SENSE Agenda Items for SCSI-2 Bold text below indicates action items. 1. Additional Sense Code resolution [88-032] Considerable time was spent reviewing 88-032R0 and preparing 88-032R1. All significant conflicts were resolved with minimal impact to existing products. There are still a few questions left to be resolved with the scanner group. The working group recommends that 88-032R1 be included in S2R5. The working group requests that any disagreements with 88-032R1 be promptly exposed in writing so that this issue can be put behind us. 2. [] bracket removal 3. Review of S2R4 These two agenda items were covered together late in the meeting. The July working group should devote much more attention to these two items. Page 4-14: Vterm needs to be defined for the B cable. Bob Snively will prepare a proposal on these numbers. John Lohmeyer suggested that the B cable terminations be powered (as opposed to passive) in anticipation of the "more than 8 devices" proposal which is in the queue for SCSI-3. John Lohmeyer will prepare a proposal on this. Page 5-8 [] note on relationship of REQ and REQB: Bob Snively stated that more text is needed on the relationship of these two signals. He will prepare a proposal on this. The editors should create a section 5.5.2 that describes one-byte, two-byte, and multi-byte (extended) messages. 5.5.3 should then document all messages in alphabetical order. Page 5-19: The CLEAR QUEUE message description (and the other queuing messages) varies from the normal message description in style. The editors should "normalize" the style. Page 5-21: "Logical thread" is not the proper term to use in the queuing descriptions. We should either use "initiator-logical unit pair" or "virtual circuit" or "path" or some other phrase that we define and place in the glossary. Page 5-28: The [] note will be resolved by adding 88-37, which was accepted by the April plenary meeting. Page 6-2: No one in the working group understood the problem stated in the [] note. The editors should delete the [] note unless Larry Lamers can provide an explanation of the problem. Page 6-6: The cache control bits were deleted, not moved. There was some question of whether they should be put back in Section 6 or 8. Gerry Houlder wanted them to be restored to section 6 because they apply to more that the Direct-Access Device command set. Others agreed with the move to section 8 in that these bits do not apply to many device types. The editors will restore the cache control bits. (The definition may change slightly if 88-030 is accepted.) Page 6-8: Delete the [] note. The working group feels that the VU bits in the control byte should remain. Page 6-15: Delete the [] note. Section 6: A section on tagged queuing from 87-059R4 which was intended to be placed in section 6 is not there. The editors should include it. Page 6-18: Larry Lamers and Bob Snively need to resolve their war. Some working group members questioned whether it is really necessary to be able to enter an ECA state via AEN. Page 7-2: Delete the [] note. Page 7-4: We should simplify the CHANGE DEFINITION command to require the definition to be consistent across all logical units on the target. The [] should stay in until resolved. Page 7-10: Do we really need a proposal for COPYing from a scanner? Delete the [] note. COPY managers can refuse to copy from a scanner until/unless we get a definitive description. (Two of the last four people had to leave the meeting at this point, so we adjourned...) The remainder of the document should be on the July working group's agenda. 4. Rev 4 Issues - Sections 7 and 8 [88-38, 88-57] (J. McGrath) Point 1: The paragraph from REQUEST SENSE listing the fatal errors that can cause CHECK CONDITION should also be included in INQUIRY. Point 2: The question of whether the target should respond to a request for an unimplemented page with CHECK CONDITION status and ILLEGAL REQUEST sense key or return a zero page length arose again. Most agreed that we should not permit two valid mechanisms, but there was some concern that both implementations exist in practice. Jim McGrath investigated the "prior practice" on unsupported features over Tuesday evening and returned on Wednesday with 88-57 that documents this practice. In most cases, unsupported fields and bits cause CHECK CONDITION status and ILLEGAL REQUEST sense key. The working group recommends that CHECK CONDITION status and ILLEGAL REQUEST sense key be the only permitted method of responding to a request for an unimplemented page. Jim McGrath will revise 88-57 to suggest the new wording. Point 3: No change Point 4: "Ignore" really means ignore. The editors will change the wording in the reserve commands concerning which commands pass through a reservation to be "... an INQUIRY command or a REQUEST SENSE command (which shall be executed) or a ...". Point 5: "Ignore" really means ignore. Point 6: Not a valid issue. Point 7: WRITE SAME needs to be fixed! During the meeting, Jeff Stai revised the WRITE SAME command and provided it to John Lohmeyer who will forward it to Larry Lamers. 5. WORM Device Specific MODE SELECT Page [88-44] (P. Boulay) Deferred -- Paul not present. 6. Pin assignments for high density connector B Nothing was done by the working group on this item. John Lohmeyer said he is waiting for Ed Gardner to complete his action item to prepare revision 1 of 88-048. 7. ACCESS LOG proposals (Spence) [88-006, 88-053] 88-053: There was general agreement on the parameters proposed by Bill Spence. There was a question of who "turns the crank." We may need to add a page on seek parameters. Do the parameters need a description or is the name sufficient definition? Most thought the name was sufficient for the purposes of the standard. 88-006: (Jim McGrath's LOG SELECT/SENSE proposal). The parameter list length is to be shortened to two bytes in bytes 7&8 to be consistent with other CDBs. The operation codes are to be selected by editors. 88-006 allows parameters from 1-255 bytes in length. Should counters also be variable in length or fixed or a small number of discreet values? The conclusion was to be variable, but add an implementors note recommending certain lengths (1,2,4,8). Dal's KISS proposal: Reduce complexity by using a single-counter philosophy: The target reports whether the counters are volatile or non- volatile. Keep the threshold values, but the saved values go away. Jim McGrath will revise his proposal before the next plenary meeting. 8. Long vs. Short BUSY [87-192R3] (Semenak) Jim prepared yet another revision of his proposal which uses NOT READY sense key to indicate that the target is in a CA or ECA state. This would only apply to targets that support per-initiator sense data. Dal objected that Jim's proposal would require targets to report CHECK CONDITION status to all other initiators, even if the BUSY condition is expected to be short. Bob Snively objected that the Jim's proposal really didn't solve the basic problem in that even knowing who has the target BUSY won't get it un-BUSY any faster. The discussion did not converge -- The working group asked Jim and his foes to take issue off-line to resolve their differences. This issue will remain on the plenary agenda for resolution. 9. RECOVER BUFFERED DATA command for sequential access devices [88-022] (Appleyard) Deferred -- Dennis not present. 10. Residues on WRITE FILEMARKS commands (Davies) This item refers to the [] note on page 9-41. The method of reporting residues doesn't work right since we added the Immed bit to the WRITE FILEMARKS command. Dan Davies will correct the filemark residue reporting in section 9. 11. MODE SELECT rewording on saving pages and nonchangeable parameters [88- 027] (Houlder) Those parts of Gerry's proposal that were not accepted at the March working group were reviewed. The working group0 agreed to add the second to last paragraph of page 1 of 88-027 to page 7-35 (first full paragraph). On page 7-35, "(1) round the value..." should become "(1) round the parameter...". The editors should add a sentence/paragraph describing that the initiator should use the same page lengths that the target reports it supports in MODE SENSE. If the initiator sends a different length, the target shall report CHECK CONDITION status, ILLEGAL REQUEST sense key, and an ASC of INVALID PARAMETER IN PARAMETER LIST. 12. Proposed definition of caching [88-030] (McGrath) The working group reviewed 88-030R0. A number of editorial corrections were suggested. Jim agreed to make these changes in revision 1 of 88-030. The working group recommends that 88-030R1 be accepted in S2R5. 13. Wide Data Transfer Request (Goldman) Deferred pending input from Steve Goldman. 14. Embedded Disk Cache parameters [88-049] (McGrath) The embedded disk cache SIWG had met on Monday evening at St. Pete. Steve Goldman was expected to send us a marked-up version of 88-049. He didn't. Dal explained what he understood the agreement to be. The minimum and maximum prefetch lengths were being altered to select absolute or multiple via a bit rather than having two fields each -- one of which is normally zero. Jim McGrath brought copies of rev 1 of this proposal on Tuesday. Jim reviewed the differences in rev 1. Several improvements were suggested by the working group. The priority fields will be reduced from 8 bits to 2 or 3 bits (to be determined by Jim). Issue: Should we add a bit to control write-back or write-through for write commands? Jim argues that this should be in a new page for write cache control parameters. He agreed to add a one-bit page to his proposal. Jim McGrath will prepare a revision 2 of 88-049 that includes the working group's comments for discussion at the next plenary meeting. The working group recommends that 88-049R2 be included in S2R5. 15. Sequential Access Initialization Procedure (88-42) This document was briefly reviewed and a number of minor enhancements were made. The Working Group recommends that 88-042, as revised, be included in S2R5. 16. MODE SELECT Clarification (88-43) Dan Davie's active format field was revised as follows: 00h Default 01h-09h Use 01h-09h tracks at a time 0Ah Use 18 tracks at a time 0Bh-1Fh Reserved Minor changes were made to the second part of Dan's proposal to make it clearer that the medium format recognition capability is a function of both the target's and the medium's capabilities. Dan Davies will prepare a rev 1 of 88-43 with the agreed to changes. The working group recommends that 88-043R1 be included in S2R5. 17. "User-friendly" MODE SELECT commands It was generally agreed that a mechanism should be present to permit permissivism. Most of the discussion was over whether one or two bits should be used in specifying permissive vs. strict parameter interpretation. Do we permit the 00b to be target-specific? 00b Target-Specific Action 01b Strict interpretation of MODE SELECT parameters 10b Lenient interpretation of MODE SELECT parameters 11b Reserved Dal advocated that these bits should be in the page headers. Target manufacturers objected to this. After a lot of discussion (and the realization how much all of this would cost in terms of implementing code and documenting the standard), it was agreed that the first implementors note on page 7-43 more-or-less describes what is really needed, but it is too general. See 87-167R1 where it came from. Also, it should become part of the document, not an implementors note. The "two-bit" non-proposal is dead. Instead, if a MODE SELECT is issued with a new block size and no pages, then the target will be permitted to calculate the correct page parameters from the desired block length. Similar solutions will be defined for other interdependent parameters as they are identified. Action Item: Jim Semenak to "take care of it". 18. Helical Scan Density Code [88-52] (Hess) The working group recommends that 88-52 be accepted and included in Rev 5. 19. Peter Johansson's Suggestion re SDTR Messages Some minor revision to John Lohmeyer's alternative wording was accepted. He will revise his wording to reflect the revisions and assign a document number. The working group recommends that this document be included in S2R5. 20. Wednesday Lunch After considerable debate, the working group agreed to call out for pizza for lunch on Wednesday. Agenda Items for SCSI-3 The below five agenda items were not addressed due to lack of time. 50. SEND DIAGNOSTIC Pages [87-186] (Spence) 51. LOAD SKIP MASK command proposal [87-203] [87-217] (Floryance, McIntyre) 52. More that 8 devices on wide SCSI [87-206] (Harms) 53. Search Command modifications [88-002] (Stai) 54. Expanded RelAdr Bit [88-007] (Boulay) Action Items: In addition to the action items shown in bold text above, the following action items were identified: Jim McGrath will put his revised caching proposal on the SCSI BBS. The editors will define "Peripheral Device" more clearly in glossary. The editors will define "invalid" in the glossary to mean illegal (i.e., usage of a reserved field or code) or unsupported (option not implemented). The editors will add a note to appendix D saying that the old alternative 1 shielded connector from X3.131-1986 has been deleted in SCSI-2. Dan Davies will correct the filemark residue reporting in section 9. The editors will restore the definitions of the cache control bits. Rev 4 says that they were moved from section 6 to section 8, but they do not appear in section 8. (These bit definitions are slightly different in 88- 030R1 which the working group has recommended be included in rev 5.) The working group recommends that the term "nonchangeable" be replaced with "target-defined".